WO2023143570A1 - 一种连接关系预测方法及相关设备 - Google Patents

一种连接关系预测方法及相关设备 Download PDF

Info

Publication number
WO2023143570A1
WO2023143570A1 PCT/CN2023/073707 CN2023073707W WO2023143570A1 WO 2023143570 A1 WO2023143570 A1 WO 2023143570A1 CN 2023073707 W CN2023073707 W CN 2023073707W WO 2023143570 A1 WO2023143570 A1 WO 2023143570A1
Authority
WO
WIPO (PCT)
Prior art keywords
network
network device
information
connection relationship
target
Prior art date
Application number
PCT/CN2023/073707
Other languages
English (en)
French (fr)
Inventor
周敏
李必盛
武可
黄增峰
Original Assignee
华为技术有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 华为技术有限公司 filed Critical 华为技术有限公司
Publication of WO2023143570A1 publication Critical patent/WO2023143570A1/zh

Links

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/12—Discovery or management of network topologies
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/06—Management of faults, events, alarms or notifications
    • H04L41/0631—Management of faults, events, alarms or notifications using root cause analysis; using analysis of correlation between notifications, alarms or events based on decision criteria, e.g. hierarchy, tree or time analysis
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/14—Network analysis or design
    • H04L41/147—Network analysis or design for predicting network behaviour

Definitions

  • the present application relates to the field of artificial intelligence, in particular to a connection relationship prediction method and related equipment.
  • Artificial intelligence is a theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.
  • artificial intelligence is the branch of computer science that attempts to understand the nature of intelligence and produce a new class of intelligent machines that respond in ways similar to human intelligence.
  • Artificial intelligence is to study the design principles and implementation methods of various intelligent machines, so that the machines have the functions of perception, reasoning and decision-making.
  • Network topology is the most basic and extensive network information basis for service engineers. Accurate, complete, and real-time service path topology restoration is very competitive in the field of operation and maintenance. Superimposed information such as alarms and performance on the topology module can quickly realize fault diagnosis. Topology is also the most basic element of network visualization. However, the current topology is incomplete and inaccurate in various sites and scenarios. Taking the telecommunication network as an example, the telecommunication network can include network domains such as core network, data bearer network, transmission network, wireless access network, and fixed access network.
  • the current telecommunication network can generate a network topology diagram of a single network domain, for example , the network elements in a single network domain are configured by the network management system in the single network domain, so the network management system in the single network domain can generate the network topology information of the single network domain according to the configuration information recorded when configuring the network elements, the topology
  • the information is used to describe the connection relationship between network elements in a single network domain.
  • Figure 1 is a schematic diagram of the corresponding scene. Using this method to generate an overall network topology map of multiple network domains has low efficiency and low accuracy, and cannot meet user needs.
  • connection relationship prediction method which will carry the first network information related to the topology structure information of the first network device and the second network device, and carry the operation information related to the first network device and the second network device.
  • the second network information of the state information is used as a reference to predict whether there is a communication connection relationship between the first network device and the second network device.
  • connection relationship prediction method the method comprising:
  • the target network includes a plurality of network devices, the plurality of network devices include a first network device and a second network device, and the first network information includes topology information of the communication link where the first network device and the second network device are located, and the second network information packet Operating status information of multiple network devices including the first network device and the second network device;
  • the first network device and the second network device may be any two network devices in the target network.
  • the first network device and the second network device may be designated by the user.
  • the user may specify to predict the communication connection relationship between which two network devices in the target network.
  • first network information of the target network may be acquired, where the first network information may include communication connection relationships between multiple network devices in the target network.
  • a communication connection relationship between network devices may be understood as: there is a physical communication path (such as a wired communication path or a wireless communication path, etc.) between network devices, or there is traffic interaction between network devices.
  • An embodiment of the present application provides a method for predicting a connection relationship, the method comprising: acquiring first network information and second network information of a target network; wherein the target network includes a plurality of network devices, and the plurality of network devices including a first network device and a second network device, the first network information includes topology information of a communication link where the first network device and the second network device are located, and the second network information includes the The operating status information of multiple network devices including the first network device and the second network device; according to the first network information and the second network information, through feature extraction, the first network device and the first network device are obtained.
  • a target feature vector of the second network device predicting a communication connection relationship between the first network device and the second network device through a target neural network according to the target feature vector.
  • This application regards the first network information carrying the topology information related to the first network device and the second network device, and the second network information carrying the operating status information related to the first network device and the second network device as the first network information for prediction.
  • the reference of whether there is a communication connection relationship between the network device and the second network device can identify richer data association relationships when predicting the connection relationship between network devices, thereby increasing the prediction accuracy of the connection relationship.
  • the alarm information includes at least one of the following information: a time when an alarm occurs on the network device, and an alarm type that occurs on the network device.
  • the alarm information may be an alarm record of a network device, and the target network includes multiple devices, and each alarm record may include a name of the alarm, a time when the alarm occurs, and an identifier of the network device where the alarm occurs.
  • the name of the alarm may also be referred to as the type of the alarm; there may be various forms of identification of the device where the alarm occurs, which is not specifically limited in this embodiment of the present application.
  • the identification of the device where the alarm occurs may be a serial number.
  • alarm records usually exist in an alarm log. Therefore, the alarm record can be obtained based on the alarm log of the target network.
  • alarm records generated by the target network within a target time period may be obtained, where the target time period may be set according to actual needs.
  • the target time period may be one month, in addition, the target time period may also be 20 days, 25 days, 35 days, 40 days and so on.
  • the key performance indicator of the network device may be the communication flow of the network device.
  • the obtaining the target feature vectors of the first network device and the second network device through feature extraction according to the first network information and the second network information includes:
  • multiple embedding vectors of multiple network devices including the first network device and the second network device are obtained; wherein, the network device whose operating status information has a greater similarity corresponds to The closer the distance between the embedding vectors; the multiple embedding vectors include the first embedding vector of the first network device and the second embedding vector of the second network device;
  • the running status information can be mapped to the space of embedding vectors, that is, a higher-dimensional feature space, and then the running status information can be expressed in the form of embedding vectors, where the purpose of mapping can be is: the similarity of the corresponding embedding vectors between network devices with higher similarity of operating state information is higher.
  • the multiple network devices including the first network device and the second network device include: at least one network device other than the first network device and the second network device Internet equipment.
  • the multiple network devices including the first network device and the second network device may include all network devices included in the target network.
  • the obtaining multiple embedding vectors of multiple network devices including the first network device and the second network device according to the second network information includes:
  • the second network information obtain the semantic topology of multiple network devices including the first network device and the second network device, where the semantic topology includes semantic connection relationships between network devices, where the The semantic connection relationship exists between the network devices whose operating state information has a similarity greater than a threshold;
  • a random walk is performed on the semantic topology to obtain multiple embedding vectors of multiple network devices including the first network device and the second network device.
  • the first network device and the second network device are obtained through feature extraction according to the first network information, the first embedding vector, and the second embedding vector
  • the target feature vector including: according to the first network information, the semantic topology information, the first embedding vector and the second embedding vector, through feature extraction, obtain the first network device and the Target feature vector for the second network device.
  • the acquiring the first network information and the second network information of the target network includes: acquiring the first network information and the second network information of the target network from the user equipment;
  • the method further includes: after predicting the communication connection relationship between the first network device and the second network device, connecting the communication connection between the first network device and the second network device The relationship is communicated to the user device.
  • the feature extraction is implemented based on a feature extraction network, and the feature extraction network is a graph neural network GNN; the target neural network is a fully connected network.
  • connection relationship prediction method the method comprising:
  • the target network includes a plurality of network devices
  • the first network information includes a topology of communication connections among the plurality of network devices information
  • the second network information includes operating status information of the multiple network devices
  • the multiple network devices include a first network device and a second network device
  • the running status information includes at least one of the following information:
  • the alarm information includes at least one of the following information:
  • the time when the network device generates an alarm, and the type of alarm that occurs on the network device is the time when the network device generates an alarm, and the type of alarm that occurs on the network device.
  • the method also includes:
  • the third network information includes the third network information and the multiple The communication connection relationship of at least one network device among the network devices, the fourth network device includes the operation status information of the third network device;
  • the fourth network device predicting a communication connection relationship between the third network device and at least one network device among the plurality of network devices
  • the present application provides a device for predicting a connection relationship, the device comprising:
  • An acquisition module configured to acquire first network information and second network information of a target network; wherein, the target network includes a plurality of network devices, and the plurality of network devices include a first network device and a second network device, the The first network information includes topology information of the communication link where the first network device and the second network device are located, and the second network information includes the first network device and the second network device The operating status information of multiple network devices;
  • a feature extraction module configured to obtain target feature vectors of the first network device and the second network device through feature extraction according to the first network information and the second network information;
  • a connection relationship prediction module configured to predict the communication connection relationship between the first network device and the second network device through a target neural network according to the target feature vector.
  • the running status information includes at least one of the following information:
  • the alarm information includes at least one of the following information:
  • the time when the network device generates an alarm, and the type of alarm that occurs on the network device is the time when the network device generates an alarm, and the type of alarm that occurs on the network device.
  • the feature extraction module is specifically used for:
  • multiple embedding vectors of multiple network devices including the first network device and the second network device are obtained; wherein, the network device whose operating status information has a greater similarity corresponds to The closer the distance between the embedding vectors; the multiple embedding vectors include the first embedding vector of the first network device and the second embedding vector of the second network device;
  • the multiple network devices including the first network device and the second network device include:
  • the feature extraction module is specifically used for:
  • the second network information obtain the semantic topology of multiple network devices including the first network device and the second network device, where the semantic topology includes semantic connection relationships between network devices, where the The semantic connection relationship exists between the network devices whose operating state information has a similarity greater than a threshold;
  • a random walk is performed on the semantic topology to obtain multiple embedding vectors of multiple network devices including the first network device and the second network device.
  • the feature extraction module is specifically used for:
  • the acquiring module is specifically used for:
  • the device also includes:
  • a sending module configured to send the communication connection relationship between the first network device and the second network device to to the user device.
  • the feature extraction is implemented based on a feature extraction network, and the feature extraction network is a graph neural network GNN; the target neural network is a fully connected network.
  • the present application provides a device for predicting a connection relationship, the device comprising:
  • An acquisition module configured to receive first network information and second network information of a target network from a user equipment; wherein the target network includes a plurality of network devices, and the first network information includes information between the plurality of network devices The topology information of the communication connection, the second network information includes the operation state information of the plurality of network devices; the plurality of network devices include a first network device and a second network device;
  • a connection relationship prediction module configured to predict a communication connection relationship between the first network device and the second network device according to the first network information and the second network information;
  • a sending module configured to transmit the communication connection relationship between the first network device and the second network device to the user equipment.
  • the running status information includes at least one of the following information:
  • the alarm information includes at least one of the following information:
  • the time when the network device generates an alarm, and the type of alarm that occurs on the network device is the time when the network device generates an alarm, and the type of alarm that occurs on the network device.
  • the acquisition module is also used to:
  • the third network information includes the third network information and the multiple The communication connection relationship of at least one network device in the network devices, the fourth network device includes the operating status of the third network device information;
  • the connection relationship prediction module is further configured to: predict the third network device and the fourth network device according to the first network information, the second network information, the third network information, and the fourth network device.
  • the sending module is further configured to: transfer the communication connection relationship between the third network device and at least one network device among the plurality of network devices to the user equipment.
  • connection relationship prediction device which may include a memory, a processor, and a bus system, wherein the memory is used to store programs, and the processor is used to execute the programs in the memory to perform the above-mentioned first Aspect and any optional method thereof, second aspect and any optional method thereof.
  • the embodiment of the present application provides a computer-readable storage medium, where a computer program is stored in the computer-readable storage medium, and when it is run on a computer, the computer executes the above-mentioned first aspect and any one thereof.
  • the embodiment of the present application provides a computer program, which, when running on a computer, enables the computer to execute the above-mentioned first aspect and any optional method thereof, the second aspect and any optional method thereof .
  • the present application provides a system-on-a-chip, which includes a processor, configured to support the model distillation device to implement the functions involved in the above-mentioned aspect, for example, to send or process the data involved in the above-mentioned method; or, information.
  • the chip system further includes a memory, and the memory is used for storing necessary program instructions and data of the execution device or the training device.
  • the system-on-a-chip may consist of chips, or may include chips and other discrete devices.
  • An embodiment of the present application provides a method for predicting a connection relationship, the method comprising: acquiring first network information and second network information of a target network; wherein the target network includes a plurality of network devices, and the plurality of network devices including a first network device and a second network device, the first network information includes topology information of a communication link where the first network device and the second network device are located, and the second network information includes the The operating status information of multiple network devices including the first network device and the second network device; according to the first network information and the second network information, through feature extraction, the first network device and the first network device are obtained.
  • a target feature vector of the second network device predicting a communication connection relationship between the first network device and the second network device through a target neural network according to the target feature vector.
  • This application regards the first network information carrying the topology information related to the first network device and the second network device, and the second network information carrying the operating status information related to the first network device and the second network device as the first network information for prediction.
  • the reference of whether there is a communication connection relationship between the network device and the second network device can identify richer data association relationships when predicting the connection relationship between network devices, thereby increasing the prediction accuracy of the connection relationship.
  • Fig. 1 is a kind of structural schematic diagram of main frame of artificial intelligence
  • Fig. 2 is a schematic diagram of a network structure
  • Fig. 3 is a schematic diagram of a network structure
  • FIG. 4 is a schematic diagram of an application architecture
  • Fig. 5 is a schematic flow chart of a connection relationship prediction method
  • FIG. 6 is a schematic diagram of first network information
  • Figure 7 is a schematic illustration of a node label
  • Fig. 8 is a schematic flow chart of encoding the first network information
  • Figure 9 is a schematic diagram of the construction of a semantic topology
  • Figure 10 is a schematic diagram of building an embedding vector based on a random walk
  • Figure 11 is a schematic diagram of feature fusion
  • FIG. 12 is a schematic diagram of a model training process provided by the embodiment of the present application.
  • FIG. 13 is a schematic flowchart of a method for predicting a connection relationship provided in an embodiment of the present application.
  • FIG. 14 is a schematic structural diagram of a connection relationship prediction device provided by an embodiment of the present application.
  • FIG. 15 is a schematic structural diagram of a connection relationship prediction device provided by an embodiment of the present application.
  • Fig. 16 is a schematic structural diagram of the execution device provided by the embodiment of the present application.
  • Fig. 17 is a schematic structural diagram of a training device provided by an embodiment of the present application.
  • FIG. 18 is a schematic structural diagram of a chip provided by an embodiment of the present application.
  • Figure 2 shows a schematic structural diagram of the main framework of artificial intelligence.
  • the following is from the “intelligent information chain” (horizontal axis) and “IT value chain” ( Vertical axis) to illustrate the above artificial intelligence theme framework in two dimensions.
  • the "intelligent information chain” reflects a series of processes from data acquisition to processing. For example, it can be the general process of intelligent information perception, intelligent information representation and formation, intelligent reasoning, intelligent decision-making, intelligent execution and output. In this process, the data has undergone a condensed process of "data-information-knowledge-wisdom".
  • IT value chain reflects the value brought by artificial intelligence to the information technology industry from the underlying infrastructure of artificial intelligence, information (provided and processed by technology) to the systematic industrial ecological process.
  • the infrastructure provides computing power support for the artificial intelligence system, realizes communication with the outside world, and realizes support through the basic platform.
  • the basic platform includes distributed computing framework and network and other related platform guarantees and supports, which can include cloud storage and computing, interconnection and interworking networks, etc.
  • sensors communicate with the outside to obtain data, and these data are provided to the smart chips in the distributed computing system provided by the basic platform for calculation.
  • Data from the upper layer of the infrastructure is used to represent data sources in the field of artificial intelligence.
  • the data involves graphics, images, voice, text, and IoT data of traditional equipment, including business data of existing systems and sensory data such as force, displacement, liquid level, temperature, and humidity.
  • Data processing usually includes data training, machine learning, deep learning, search, reasoning, decision-making, etc.
  • machine learning and deep learning can symbolize and formalize intelligent information modeling, extraction, preprocessing, training, etc. of data.
  • Reasoning refers to the process of simulating human intelligent reasoning in a computer or intelligent system, and using formalized information to carry out machine thinking and solve problems according to reasoning control strategies.
  • the typical functions are search and matching.
  • Decision-making refers to the process of decision-making after intelligent information is reasoned, and usually provides functions such as classification, sorting, and prediction.
  • some general capabilities can be formed based on the results of data processing, such as algorithms or a general system, such as translation, text analysis, computer vision processing, speech recognition, image processing identification, etc.
  • Intelligent products and industry applications refer to the products and applications of artificial intelligence systems in various fields. It is the packaging of the overall solution of artificial intelligence, which commercializes intelligent information decision-making and realizes landing applications. Its application fields mainly include: intelligent terminals, intelligent transportation, Smart healthcare, autonomous driving, smart cities, etc.
  • FIG. 3 is a schematic diagram of the architecture of a network system (such as the target network in the embodiment of the application) provided by the embodiment of the application.
  • the network system may include a first node 201, a second node 202, and a plurality of Network domain 203, the plurality of network domains 203 may include two or more network domains in the core network, data bearer network, transport network, wireless access network, and fixed access network, for example, including two of the network domains , or 3 domains, or 4 domains, or 5 domains, etc.
  • Each network domain 203 includes a plurality of network elements 204 (network elements in this embodiment of the application may also be referred to as network devices, such as a first network device, a second network device, and a third network device).
  • network elements in this embodiment of the application may also be referred to as network devices, such as a first network device, a second network device, and a third network device.
  • the specific number of network elements included in the unit 204 is not limited here; FIG. 2 only illustrates the network elements in one network domain, and the structures in other network domains can be deduced in the same way.
  • the core network may include a series of network elements such as packet switched (PS), circuit switched (CS), and home subscriber server (HSS), and the data bearer network may include switches, routers, firewalls, etc.
  • PS packet switched
  • CS circuit switched
  • HSS home subscriber server
  • the transport network can include microwave, multi-service transmission platform (multi-service transmission platform, MSTP), wavelength division, packet transport network (packet transport network, PTN) and other types of network elements
  • the wireless access network can include the first The second-generation mobile communication technology (The 2nd-Generation, 2G), the third-generation mobile Communication technology (The 3rd-Generation, 3G), the fourth-generation mobile communication technology (the 4th Generation mobile communication, 4G), the base station and the base station controller in the fifth-generation mobile communication technology (the 5th-Generation, 5G)
  • the fixed access network can include a series of network elements such as optical line terminal (OLT), optical network terminal (ONT), and multiple x unit (MxU). etc.
  • the first node 201 can be a user equipment such as a terminal device or a server, and the first node 201 can obtain the network topology and operation status information of the network system, and can use the network topology (indicating the communication connection relationship between network elements) And the running status information is transmitted to the second node 202, the second node 202 can be a server or a terminal device, and the second node 202 can provide the first node 201 with the right and wrong judgments of the connection relationship between the network elements in the network system and the complete information.
  • the second node 202 can judge whether the communication connection relationship between the network elements indicated by the network topology uploaded by the first node 101 is accurate or whether there is omission according to the information uploaded by the first node 201, and The judgment result is transmitted to the first node 101 .
  • the second node 202 may provide the first node 201 with the prediction service of the above connection relationship through an application programming interface (application programming interface, API).
  • application programming interface application programming interface
  • the second node 202 may be a terminal device or a device with cloud computing capabilities, and an application program for providing the prediction service of the above connection relationship may be installed on the second node 202, through interaction with the first node 201 to provide The first node 201 provides the prediction service of the above connection relationship.
  • the neural network can be composed of neural units, and the neural unit can refer to an operation unit that takes xs (that is, the input data) and an intercept 1 as input, and the output of the operation unit can be:
  • Ws is the weight of xs
  • b is the bias of the neuron unit.
  • f is the activation function of the neural unit, which is used to introduce nonlinear characteristics into the neural network to convert the input signal in the neural unit into an output signal.
  • the output signal of the activation function can be used as the input of the next convolutional layer, and the activation function can be a sigmoid function.
  • a neural network is a network formed by connecting multiple above-mentioned single neural units, that is, the output of one neural unit can be the input of another neural unit.
  • the input of each neural unit can be connected with the local receptive field of the previous layer to extract the features of the local receptive field.
  • the local receptive field can be an area composed of several neural units.
  • Deep Neural Network also known as multi-layer neural network
  • DNN Deep Neural Network
  • the neural network inside DNN can be divided into three categories: input layer, hidden layer, and output layer.
  • the first layer is the input layer
  • the last layer is the output layer
  • the layers in the middle are all hidden layers.
  • the layers are fully connected, that is, any neuron in the i-th layer must be connected to any neuron in the i+1-th layer.
  • the coefficient of the kth neuron of the L-1 layer to the jth neuron of the L layer is defined as It should be noted that the input layer has no W parameter.
  • more hidden layers make the network more capable of describing complex situations in the real world. Theoretically speaking, a model with more parameters has a higher complexity and a greater "capacity", which means that it can complete more complex learning tasks.
  • Training the deep neural network is the process of learning the weight matrix, and its ultimate goal is to obtain the weight matrix of all layers of the trained deep neural network (the weight matrix formed by the vector W of many layers).
  • a graph is a data structure comprising at least one vertex and at least one edge.
  • the vertices in the graph can be mapped to entities, and the edges in the graph can be mapped to the relationship between entities.
  • a graph can be directed or undirected.
  • a graph can also include other data other than vertices and edges, such as labels of vertices and labels of edges.
  • each vertex in the graph can represent a user, and each edge in the graph can represent the social relationship between different users, and the data of each vertex in the graph is User portrait data and user behavior data, such as the user's age, occupation, hobbies, education, etc.
  • each vertex in the graph can represent a user or a product, and each edge in the graph can represent the interactive relationship between the user and the product, such as purchase relationship, collection relationship, etc. .
  • each vertex in the graph can represent an account, transaction or fund.
  • the edges in the graph can represent the flow relationship of funds, for example, the loops in the graph can represent circular transfers. For another example, it is applied to the scene of determining the connection relationship between network elements in the network system.
  • Each vertex in the graph can represent a network element, such as a router, switch, terminal, etc., and each edge in the graph can represent a different network element. connection relationship between.
  • a subgraph is a part of the graph, including some vertices and some edges in the graph.
  • a subgraph can also be called a partition in the graph (English: partition).
  • a graph can contain multiple subgraphs.
  • GNN is a deep learning method with structural information that can be used to calculate the current state of a node.
  • the information transmission of the graph neural network is carried out according to the given graph structure, and the state of each node can be updated according to the adjacent nodes. Specifically, it can use the neural network as the aggregation function of point information according to the structure diagram of the current node to transfer the information of all adjacent nodes to the current node, and update it in combination with the state of the current node.
  • the output of the graph neural network is the state of all nodes.
  • the convolutional neural network can use the back propagation (BP) algorithm to correct the size of the parameters in the initial super-resolution model during the training process, so that the reconstruction error loss of the super-resolution model becomes smaller and smaller. Specifically, passing the input signal forward until the output will generate an error loss, and updating the parameters in the initial super-resolution model by backpropagating the error loss information, so that the error loss converges.
  • the backpropagation algorithm is a backpropagation movement dominated by error loss, aiming to obtain the parameters of the optimal super-resolution model, such as the weight matrix.
  • FIG. 4 is a schematic diagram of a system architecture provided by an embodiment of the present application.
  • the system architecture 500 includes an execution device 510 , a training device 520 , a database 530 , a client device 540 , a data storage system 550 and a data collection system 560 .
  • the execution device 510 includes a calculation module 511 , an I/O interface 512 , a preprocessing module 513 and a preprocessing module 514 .
  • the calculation module 511 may include the target model/rule 501, and the preprocessing module 513 and the preprocessing module 514 are optional.
  • the data collection device 560 is used to collect training samples.
  • the training samples may be image data, text data, audio data, etc.
  • the training samples are topology information of a network (such as a target network) and operation status information of network elements. After collecting the training samples, the data collection device 560 stores these training samples in the database 530 .
  • the training device 520 can obtain the target model/rule 501 based on the training samples maintained in the database 530 and the neural network to be trained (such as the feature extraction network and the target neural grid in the embodiment of the present application).
  • the training samples maintained in the database 530 are not necessarily all collected by the data collection device 560, and may also be received from other devices.
  • the training device 520 does not necessarily perform the training of the target model/rule 501 based entirely on the training samples maintained by the database 530, and it is also possible to obtain training samples from the cloud or other places for model training. Limitations of the Examples.
  • the target model/rule 501 trained according to the training device 520 can be applied to different systems or devices, such as the execution device 510 shown in FIG. 4 , which can be a terminal, such as a mobile phone terminal, a tablet computer, a notebook Computers, augmented reality (augmented reality, AR)/virtual reality (virtual reality, VR) equipment, vehicle terminals, etc., can also be servers or clouds, etc.
  • the execution device 510 shown in FIG. 4 can be a terminal, such as a mobile phone terminal, a tablet computer, a notebook Computers, augmented reality (augmented reality, AR)/virtual reality (virtual reality, VR) equipment, vehicle terminals, etc., can also be servers or clouds, etc.
  • the training device 520 may deliver the trained model to the execution device 510 .
  • the execution device 510 is configured with an input/output (input/output, I/O) interface 512 for data interaction with external devices, and the user can input data to the I/O interface 512 through the client device 540 (such as this The first network information, the second network information, the third network information and the fourth network information of the target network in the embodiment of the application).
  • I/O input/output
  • the preprocessing module 513 and the preprocessing module 514 are configured to perform preprocessing according to the input data received by the I/O interface 512 . It should be understood that there may be no preprocessing module 513 and preprocessing module 514 or only one preprocessing module. When the preprocessing module 513 and the preprocessing module 514 do not exist, the calculation module 511 may be used directly to process the input data.
  • the execution device 510 When the execution device 510 preprocesses the input data, or in the calculation module 511 of the execution device 510 performs calculation and other related processing, the execution device 510 can call the data, codes, etc. in the data storage system 550 for corresponding processing , the correspondingly processed data and instructions may also be stored in the data storage system 550 .
  • the I/O interface 512 provides the processing result (such as the connection relationship between network devices in the embodiment of the present application) to the client device 540, thereby providing it to the user.
  • the user can manually specify input data, and the “manually specify input data” can be operated through the interface provided by the I/O interface 512 .
  • the client device 540 can automatically send the input data to the I/O interface 512 . If the client device 540 is required to automatically send the input data to obtain the user's authorization, the user can set the corresponding authority in the client device 540 .
  • the user can view the results output by the execution device 510 on the client device 540, and the specific presentation form may be specific ways such as display, sound, and action.
  • the client device 540 can also be used as a data collection terminal, collecting input data from the input I/O interface 512 and output results from the output I/O interface 512 as new sample data, and storing them in the database 530 .
  • the data is stored in database 530 .
  • FIG. 4 is only a schematic diagram of a system architecture provided by the embodiment of the present application, and the positional relationship between devices, devices, modules, etc. shown in the figure does not constitute any limitation.
  • the data The storage system 550 is an external memory relative to the execution device 510 , and in other cases, the data storage system 550 may also be placed in the execution device 510 . It should be understood that the above execution device 510 may be deployed in the client device 540 .
  • the calculation module 511 of the execution device 520 can obtain the code stored in the data storage system 550 to implement the connection relationship prediction method in the embodiment of the present application.
  • the calculation module 511 of the execution device 520 may include a hardware circuit (such as an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA), a general-purpose processor, digital signal processing (digital signal processing, DSP), microprocessor or microcontroller, etc.), or a combination of these hardware circuits, for example, the training device 520 can be a hardware system with the function of executing instructions, such as CPU, DSP, etc. , or a hardware system that does not have the function of executing instructions, such as ASIC, FPGA, etc., or a combination of the above-mentioned hardware systems that do not have the function of executing instructions and hardware systems that have the function of executing instructions.
  • ASIC application specific integrated circuit
  • FPGA field-programmable gate array
  • DSP digital signal processing
  • microprocessor or microcontroller etc.
  • the training device 520 can be a hardware system with the function of executing instructions, such as CPU, DSP, etc. , or a hardware system that does not have
  • the calculation module 511 of the execution device 520 may be a hardware system capable of executing instructions
  • the connection relationship prediction method provided in the embodiment of the present application may be a software code stored in a memory
  • the calculation module 511 of the execution device 520 may read from the memory Obtain the software code in, and execute the obtained software code to realize the embodiment of the present application The connection relationship prediction method provided.
  • calculation module 511 of the execution device 520 may be a combination of a hardware system that does not have the function of executing instructions and a hardware system that has the function of executing instructions.
  • the computing module 511 in the calculation module 511 is implemented by a hardware system that does not have the function of executing instructions, which is not limited here.
  • the above-mentioned training device 520 can obtain the code stored in the memory (not shown in FIG. 4, which can be integrated into the training device 520 or deployed separately from the training device 520) to realize the connection relationship in the embodiment of the present application method of prediction.
  • the training device 520 may include a hardware circuit (such as an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA), a general-purpose processor, a digital signal processor (digital signal processing, DSP), microprocessor or microcontroller, etc.), or a combination of these hardware circuits, for example, the training device 520 can be a hardware system with the function of executing instructions, such as CPU, DSP, etc., or for not A hardware system with the function of executing instructions, such as ASIC, FPGA, etc., or a combination of the above-mentioned hardware system without the function of executing instructions and a hardware system with the function of executing instructions.
  • ASIC application specific integrated circuit
  • FPGA field-programmable gate array
  • DSP digital signal processor
  • microprocessor or microcontroller etc.
  • the training device 520 can be a hardware system with the function of executing instructions, such as CPU, DSP, etc., or for not A hardware system with the function of executing instructions, such as ASIC,
  • the training device 520 may be a hardware system capable of executing instructions.
  • the connection relationship prediction method provided in the embodiment of the present application may be a software code stored in a memory, and the training device 520 may obtain the software code from the memory and execute the The acquired software code is used to implement the connection relationship prediction method provided in the embodiment of the present application.
  • the training device 520 can be a combination of a hardware system that does not have the function of executing instructions and a hardware system that has the function of executing instructions. It can be implemented by a hardware system that executes the function of the instruction, which is not limited here.
  • training devices each serving as a computing node.
  • connection relationship prediction method provided by an embodiment of the present application is introduced from the reasoning side of the model with reference to the accompanying drawings.
  • FIG. 5 is a schematic flowchart of a connection relationship prediction method provided by the embodiment of the present application.
  • the connection relationship prediction method provided by the embodiment of the present application includes:
  • first network information and second network information of a target network wherein, the target network includes multiple network devices, the multiple network devices include a first network device and a second network device, and the first network
  • the information includes topology information of the communication link where the first network device and the second network device are located, and the second network information includes a plurality of communication links including the first network device and the second network device Operating status information of network devices.
  • the user equipment can generate the topology information of the target network according to the configuration information recorded when configuring the network elements (the topology information can indicate multiple network devices included in the target network and the communication connection relationship between network devices ), however, since the topology information is collected manually, there may be inaccurate or missing connections. Therefore, the relevant data of the target network may be delivered to a device that provides a connection relationship prediction service (for example, the execution body of step 501).
  • the target network may be any communication network.
  • the embodiment of the present application does not specifically limit the scale of the target network and the topology of the target network.
  • the execution subject of step 501 may be a terminal device or a server.
  • the server may receive the first network information and the second network information of the target network sent by the user, for example, the server may receive the user's terminal information through an API or other types of interfaces provided by the cloud server.
  • the first network information and the second network information of the target network sent by the device (or server), or the cloud client installed on the server can receive the first network information and the second network information of the target network sent by the user's terminal device (or server). Second network information.
  • the client of the application program installed on the terminal device may receive the first network information and the second network information of the target network input by the user.
  • the target network includes multiple network devices (or referred to as network elements).
  • the first network information may include communication connection relationships between multiple network devices.
  • first network information of the target network may be acquired, where the first network information may include communication connection relationships between multiple network devices in the target network.
  • a communication connection relationship between network devices may be understood as: there is a physical communication path (such as a wired communication path or a wireless communication path, etc.) between network devices, or there is traffic interaction between network devices.
  • the communication connection relationship includes direct communication connection and indirect communication connection.
  • a direct communication connection can be understood as two devices can communicate directly, and an indirect communication connection can be understood as two devices need to communicate through at least one intermediate device.
  • the communication connection relationship between the base station and the microwave equipment is a direct communication connection; and the communication connection between the base station and the routing equipment is through the microwave equipment, so the communication connection relationship between the base station and the routing equipment is an indirect connection;
  • the core network devices are connected through microwave devices and routing devices, so the communication connection relationship between the base station and the core network devices is also an indirect connection.
  • the local storage based on the network center is usually used to store the information of the target network, and then the topology data of the target network (the first network information ).
  • the topology data of the target network can be obtained according to the device path log of the target network.
  • the device path log contains the data of at least one communication path, at least one communication path contains multiple devices, and each communication path contains multiple For communication-connected devices, the data of each communication path includes identifiers of multiple communication-connected devices and type information of multiple communication-connected devices.
  • the data of the communication path may include the number of the communication path, the name of the device on the communication path, the type of the device on the communication path, and the number of the device on the communication path.
  • the number of the device on the communication path can specifically indicate how many hops are required from any device on the communication path to a specific device.
  • the specific device is device 11, and accordingly, the number of device 11 on the communication path is 0; device 22 is connected to device 11, so the number of device 22 on the communication path is 1, which means that it takes one hop from device 22 to device 11; device 33 is connected to device 11 through device 22, so the number of device 33 on this communication path is 2, which means that it takes two hops from device 22 to device 11.
  • the number of the communication path, the name of the device on the communication path, the type of the device on the communication path, and the number of the device on the communication path are not limited to the forms shown in the above table.
  • the communication path numbered 1 includes device 11, device 22, and device 33, so it can be determined that there is a communication connection relationship between device 11, device 22, and device 33; the communication path numbered 2
  • the path includes the device 11 and the device 44, so it can be determined that the device 11 also has a communication connection relationship with the device 44. It can be seen that, in addition to determining the types of multiple devices in the target network according to the device path log, it is also possible to determine the communication connection relationship of multiple devices in the target network.
  • the first network information may include topology information of multiple communication links (specifically, it may be expressed as network devices on each communication link and the connection sequence between network devices), and each communication There may be a serial connection relationship between network devices on the link.
  • connection sequence between the foregoing network devices may also be described as a connection position.
  • a communication link may include network device A, network device B, and network device C, and there is a communication connection relationship between network device A and network device B, then the connection position of network device B relative to network device A is: no There are direct connections of other network devices, and network device B is separated between network device C and network device A, then network device C can be considered as the connection position relative to network device A: the connection position separated by one network device.
  • the first network information may be in the form of the above communication link topology information, or in other forms of expression, such as expressing which network devices have communication connection relationships between network devices, it should be understood , the above information may also contain the information of the communication link.
  • the first network information may include: there is a communication connection relationship between the first network device A and the network device B, there is a communication connection relationship between the network device B and the network device C, the above information It can be implicitly known that network device A, network device B, and network device C are on the same communication link, and the above information can still be derived even if there is no communication link identifier.
  • the content of the topology information of the network structure may also include other information used to describe the attributes of the network device itself, such as the identifier of the network device, the network device type, etc.
  • the information directly input by the user to the server or terminal device may not be in the form of the above-mentioned first network information, but input information describing the target network, and the server or terminal device receives the user's After the input information used to describe the topology structure of the target network, it may be preprocessed, or other information sorting methods may be used to obtain the first network information described in the above embodiment.
  • FIG. 6 is a schematic diagram of first network information, where the first network information shown in FIG. 6 may include communication link 1 and communication link 2, and communication link 1 may include network device 1 , a network device 2 and a network device 3 , the communication link 2 may include a network device 1 , a network device 4 and a network device 5 .
  • the multiple network devices include a first network device and a second network device.
  • the first network device and the second network device may be any two network devices in the target network.
  • the first network device and the second network device may be designated by the user.
  • the user may specify to predict the communication connection relationship between which two network devices in the target network.
  • the first network information includes topology information of a communication link where the first network device and the second network device are located.
  • the first network information may include topology information of the communication link where the first network device is located and topology information of the communication link where the second network device is located.
  • the topology information of the communication link where the first network device is located may be directly obtained from information directly input by the user, or obtained by deduction from information input by the user.
  • the topology information of the communication link where the second network device is located may be directly obtained from information directly input by the user, or obtained by deduction from information input by the user.
  • the topology information related to the first network device and the second network device can be obtained, that is, the above-described The topology information of the communication link where the first network device is located, and the topology structure information of the communication link where the second network device is located.
  • the topology information of the communication link where the first network device is located may be the topology information of all communication tasks of the first network device in the target network, or it may be the topology information of all communication tasks of the first network device in the target network.
  • the so-called length can be understood as the number of network devices included in the communication link, and the selection of the number of communication links and the length of the communication link may depend on the forecast The balance between accuracy and computing power, the larger the number and scale of communication links, the higher the corresponding computing power overhead, the more accurate the connection relationship obtained in the final reasoning, the smaller the number and scale of communication links, the corresponding The smaller the computational overhead, the less accurate the connection relationship obtained in the final reasoning.
  • the topology structure of the communication link where the network device to be predicted is located is taken as a consideration factor for subsequent prediction of the communication connection relationship between network devices.
  • the topology relationship needs to include the connection relationship between the network device on the communication link and the network device to be predicted, as well as the connection Location.
  • the topology information of the communication link where the first network device and the second network device are located can be encoded by a preset encoding method, and the encoding result can represent the first network device and the second network device.
  • the restoration of the physical topology can be performed.
  • the restoration of the physical topology is mainly to represent the device link relationship as a graph structure based on the existing link information.
  • the restoration of the physical topology step aims to represent the device link relationship according to the existing link information.
  • this step can be omitted for some physical topology data is already stored in the form of graph data.
  • Further structural feature encoding refers to the encoding of structural information on physical topology.
  • the physical topology restoration is to represent the device link relationship as a graph structure according to the existing link information. For each path, check whether the nodes in the path are already in the graph, and then connect the edges that should be connected in the physical topology according to Path Hop.
  • the structural feature encoding refers to the selected node pair (link) to extract a closed subgraph
  • the structure information is extracted based on the subgraph.
  • the h-order neighbor nodes h greater than or equal to 1 centered on the two nodes of this edge are extracted, as well as the edges formed between these nodes, thus forming a graph with the two nodes of this edge as A subgraph in the center is called a closed subgraph.
  • Each node in the closed subgraph is mapped to an integer set by the method of node labeling, namely:
  • the node labeling method used is derived from the following criteria:
  • the labels of the target nodes in each closed subgraph are 1 for x and y;
  • each node in the subgraph has a corresponding label, and the features of the subgraph nodes can be constructed in the way of one-hot encoding, and the feature matrix X1 of the subgraph can be obtained.
  • FIG. 7 is a schematic illustration of a node
  • FIG. 8 which is a schematic flowchart of encoding the first network information.
  • the second network information of the target network may be obtained, where the second network information includes the operation of multiple network devices including the first network device and the second network device status information.
  • the multiple network devices including the first network device and the second network device include: at least one network device other than the first network device and the second network device Internet equipment.
  • the multiple network devices including the first network device and the second network device may include all network devices included in the target network.
  • the running status information includes at least one of the following information: alarm information of the network device, and key performance indicator (key performance indicator, KPI) of the network device.
  • the alarm information includes at least one of the following information: a time when an alarm occurs on the network device, and an alarm type that occurs on the network device.
  • the alarm information may be an alarm record of a network device, and the target network includes multiple devices, and each alarm record may include a name of the alarm, a time when the alarm occurs, and an identifier of the network device where the alarm occurs.
  • the name of the alarm may also be referred to as the type of the alarm; there may be various forms of identification of the device where the alarm occurs, which is not specifically limited in this embodiment of the present application.
  • the identification of the device where the alarm occurs may be a serial number.
  • alarm records usually exist in an alarm log. Therefore, the alarm record can be obtained based on the alarm log of the target network.
  • alarm records generated by the target network within a target time period may be obtained, where the target time period may be set according to actual needs.
  • the target time period may be one month, in addition, the target time period may also be 20 days, 25 days, 35 days, 40 days and so on.
  • the key performance indicator of the network device may be the communication flow of the network device.
  • the running status information can be mapped to the space of embedding vectors, that is, a higher-dimensional feature space, and then the running status information can be expressed in the form of embedding vectors, where the purpose of mapping can be is: the similarity of the corresponding embedding vectors between network devices with higher similarity of operating state information is higher.
  • the semantic topology can be constructed based on the similarity between the running state information, similar to the above-mentioned network topology based on the communication connection relationship, the semantic topology also includes the connection relationship between multiple network devices, Different from the above-mentioned connection relationship corresponding to the communication connection, the connection relationship in the semantic topology is a relationship between network devices with relatively high similarity in operation status information.
  • FIG. 9 is a schematic diagram of the construction of a semantic topology.
  • alarm co-occurrence can be understood as the temporal similarity of alarms between network devices.
  • the specific semantic topology construction process is as follows:
  • the operation of the network device can be determined by the change characteristics of the communication flow over time similarity between states.
  • multiple embedding vectors of multiple network devices including the first network device and the second network device may be obtained according to the second network information; The distance between the embedding vectors corresponding to the network devices with greater similarity is closer; the multiple embedding vectors include the first embedding vector of the first network device and the second embedding vector of the second network device .
  • the semantic structure information can be encoded based on the semantic topology.
  • an unsupervised learning method can be used to obtain the structural feature representation of the node.
  • the embedding vector of each network device is obtained, so as to learn the structural information X2 of the semantic topology.
  • FIG. 10 is a schematic diagram of constructing an embedding vector based on a random walk.
  • the first network information may carry topology information related to the first network device and the second network device
  • the second network information may carry operating status information related to the first network device and the second network device
  • both the topology information and the running state information are related to whether there is a communication connection relationship between the first network information and the second network information.
  • the target feature vectors of the first network device and the second network device may be obtained through feature extraction according to the first network information and the second network information, and the target feature vector may be carrying the topology information related to the first network device and the second network device and the operating state information related to the first network device and the second network device, and then predicting the target feature vector through the target neural network A communication connection relationship between the first network device and the second network device.
  • the feature extraction is implemented based on a feature extraction network.
  • the feature extraction network is a graph neural network GNN.
  • the features of the first network device and the second network device can be obtained through feature extraction.
  • Target feature vector the features of the first network device and the second network device.
  • the first network device and the The target feature vector of the second network device are configured to be identical to each other.
  • the encoded first network information and the encoded second network information may be fused, and the fusion method is different from Limited to matrix stitching. Fusing the encoded first network information and the encoded second network information is equivalent to realizing a multi-source structure
  • the learned feature fusion through feature fusion, makes the fused features contain both the structural information of the physical topology and the structural information of the semantic topology. Referring to FIG. 11 , FIG. 11 is a schematic diagram of feature fusion.
  • the result of fusing the encoded first network information, the encoded second network information, and the semantic topology information is input into the feature extraction network.
  • G sub (V, E, X), wherein V represents a vertex set, E represents the edge set, X represents the fused feature matrix, its i-th row represents the feature of the i-th node in the vertex set, and A is used to represent the adjacency matrix of the subgraph.
  • V represents a vertex set
  • E represents the edge set
  • X represents the fused feature matrix
  • its i-th row represents the feature of the i-th node in the vertex set
  • A is used to represent the adjacency matrix of the subgraph.
  • Z ⁇ R N ⁇ d is the output of the graph neural network, representing the feature embedding of each subgraph.
  • Predict a communication connection relationship between the first network device and the second network device by using a target neural network according to the target feature vector.
  • the communication connection relationship between the first network device and the second network device may be predicted by using a target neural network according to the target feature vector.
  • the target neural network is a fully-connected network, wherein the fully-connected network may be a pre-trained network capable of predicting communication connection relationships between network devices according to feature vectors.
  • the output of the target neural network may be a probability that a communication connection relationship exists between the first network device and the second network device.
  • the feature vectors corresponding to any two groups of network devices in the target network can be obtained, and then the communication connection relationship between network devices can be predicted through the target neural network.
  • the final output Y can be obtained through the MLP layer, and the i-th element y i of Y represents the output obtained when the i-th image is used as an input.
  • third network information and fourth network information of the third network device from the user equipment may be received, wherein the third network device does not belong to In the target network, the third network information includes a communication connection relationship between the third network information and at least one network device among the plurality of network devices, and the fourth network device includes an operation of the third network device State information; similar to the above, the third network device and the plurality of networks may be predicted according to the first network information, the second network information, the third network information and the fourth network device a communication connection relationship between at least one network device among the devices; and delivering the communication connection relationship between the third network device and at least one network device among the plurality of network devices to the user equipment.
  • An embodiment of the present application provides a method for predicting a connection relationship, the method comprising: acquiring first network information and second network information of a target network; wherein the target network includes a plurality of network devices, and the plurality of network devices including a first network device and a second network device, the first network information includes topology information of a communication link where the first network device and the second network device are located, and the second network information includes the The operating status information of multiple network devices including the first network device and the second network device; according to the first network information and the second network information, through feature extraction, the first network device and the first network device are obtained.
  • a target feature vector of the second network device predicting a communication connection relationship between the first network device and the second network device through a target neural network according to the target feature vector.
  • This application regards the first network information carrying the topology information related to the first network device and the second network device, and the second network information carrying the operating status information related to the first network device and the second network device as the first network information for prediction.
  • the reference of whether there is a communication connection relationship between the network device and the second network device can identify richer data association relationships when predicting the connection relationship between network devices, thereby increasing the prediction accuracy of the connection relationship.
  • connection relationship prediction method in the embodiment of the present application from the reasoning process of the model, and then describes the connection relationship prediction method in the embodiment of the application from the model training process:
  • FIG. 12 is a schematic diagram of a model training process provided by the embodiment of the present application.
  • the model training provided by the embodiment of the present application includes:
  • first network information and second network information of a target network wherein, the target network includes multiple a network device, the multiple network devices include a first network device and a second network device, and the first network information includes topology information of a communication link where the first network device and the second network device are located , the second network information includes operating status information of multiple network devices including the first network device and the second network device, and the target network indicates that the first network device and the second network There is a first communication connection relationship between the devices.
  • the target network may be a network specified by the user equipment, and in the target network, there may be a communication connection relationship between the first network device and the second network device (thereby, the first network device and the second network device may be used as the model for subsequent model training Positive edge), there may not be a communication connection relationship between the first network device and the second network device (thereby, the first network device and the second network device may serve as negative edges for subsequent model training).
  • step 1201 For the specific description of step 1201, reference may be made to step 501, which will not be repeated here.
  • step 1202 For the specific description of step 1202, reference may be made to step 502, which will not be repeated here.
  • step 1203 For the specific description of step 1203, reference may be made to step 503, which will not be repeated here.
  • the communication connection relationship between the first network device and the second network device output by the target neural network, and the communication between the first network device and the second network device indicated by the target network Concatenate relations to build a loss, and train the target neural network (and feature extraction network) based on that loss.
  • a loss function needs to be set during training. Considering that this is essentially a binary classification problem, the cross-entropy loss function can be selected:
  • the cross-entropy loss function can represent the difference between the sample label and the predicted probability, so minimizing the cross-entropy loss function during the training process can keep the predicted probability of the sample consistent with the real label as much as possible.
  • FIG. 13 is a schematic flowchart of a connection relationship prediction method provided in the embodiment of the present application, the method including:
  • first network information and second network information of a target network from a user equipment; wherein the target network includes multiple network devices, and the first network information includes communication connections between the multiple network devices topology information, the second network information includes operating status information of the multiple network devices; the multiple network devices include a first network device and a second network device;
  • step 1301 For the description of step 1301, reference may be made to the description of step 501 in the foregoing embodiment, and details are not repeated here.
  • step 1302 For the description of step 1302, reference may be made to the relevant descriptions of step 502 and step 503 in the foregoing embodiments, and details are not repeated here.
  • step 1303 reference may be made to descriptions related to step 503 in the foregoing embodiments, and details are not repeated here.
  • the running status information includes at least one of the following information:
  • the alarm information includes at least one of the following information:
  • the time when the network device generates an alarm, and the type of alarm that occurs on the network device is the time when the network device generates an alarm, and the type of alarm that occurs on the network device.
  • the method also includes:
  • the third network information includes the third network information and the multiple The communication connection relationship of at least one network device among the network devices, the fourth network device includes the operation status information of the third network device;
  • the second network information, the third network information, and the fourth network device predict a relationship between the third network device and at least one network device among the plurality of network devices communication connection relationship;
  • FIG. 14 is a schematic structural diagram of a connection relationship prediction device provided in the embodiment of the present application.
  • a connection relationship prediction device 1400 provided in the embodiment of the present application includes:
  • An acquiring module 1401 configured to acquire first network information and second network information of a target network; wherein, the target network includes multiple network devices, and the multiple network devices include a first network device and a second network device, so The first network information includes the topology information of the communication link where the first network device and the second network device are located, and the second network information includes the communication link between the first network device and the second network device The operating status information of multiple network devices in the network;
  • step 501 For the description of the obtaining module 1401, reference may be made to the description of step 501 in the above embodiment, and details are not repeated here.
  • a feature extraction module 1402 configured to obtain target feature vectors of the first network device and the second network device through feature extraction according to the first network information and the second network information;
  • the connection relationship prediction module 1403 is configured to predict the communication connection relationship between the first network device and the second network device through a target neural network according to the target feature vector.
  • connection relationship prediction module 1403 For the description of the connection relationship prediction module 1403, reference may be made to the description of step 503 in the above embodiment, and details are not repeated here.
  • the running status information includes at least one of the following information:
  • the alarm information includes at least one of the following information:
  • the time when the network device generates an alarm, and the type of alarm that occurs on the network device is the time when the network device generates an alarm, and the type of alarm that occurs on the network device.
  • the feature extraction module is specifically used for:
  • multiple embedding vectors of multiple network devices including the first network device and the second network device are obtained; wherein, the network device whose operating status information has a greater similarity corresponds to The closer the distance between the embedding vectors; the multiple embedding vectors include the first embedding vector of the first network device and the second embedding vector of the second network device;
  • the multiple network devices including the first network device and the second network device include:
  • the feature extraction module is specifically used for:
  • the second network information obtain the semantic topology of multiple network devices including the first network device and the second network device, where the semantic topology includes semantic connection relationships between network devices, where the The semantic connection relationship exists between the network devices whose operating state information has a similarity greater than a threshold;
  • a random walk is performed on the semantic topology to obtain multiple embedding vectors of multiple network devices including the first network device and the second network device.
  • the feature extraction module is specifically used for:
  • the acquiring module is specifically used for:
  • the device also includes:
  • a sending module configured to send the communication connection relationship between the first network device and the second network device to to the user device.
  • the feature extraction is implemented based on a feature extraction network, and the feature extraction network is a graph neural network GNN; the target neural network is a fully connected network.
  • FIG. 15 is a schematic structural diagram of a connection relationship prediction device provided in the embodiment of the present application.
  • a connection relationship prediction device 1500 provided in the embodiment of the application includes:
  • An acquiring module 1501 configured to receive first network information and second network information of a target network from a user equipment; wherein, the target network includes a plurality of network devices, and the first network information includes one of the plurality of network devices The topology information of the inter-communication connection, the second network information includes the operation state information of the plurality of network devices; the plurality of network devices include a first network device and a second network device;
  • a connection relationship prediction module 1502 configured to predict a communication connection relationship between the first network device and the second network device according to the first network information and the second network information;
  • connection relationship prediction module 150 For the description of the connection relationship prediction module 1502, reference may be made to the description of step 1302 in the above embodiment, and details are not repeated here.
  • a sending module 1503, configured to transmit the communication connection relationship between the first network device and the second network device to the user equipment.
  • the running status information includes at least one of the following information:
  • the alarm information includes at least one of the following information:
  • the time when the network device generates an alarm, and the type of alarm that occurs on the network device is the time when the network device generates an alarm, and the type of alarm that occurs on the network device.
  • the acquisition module is also used to:
  • the third network information includes the third network information and the multiple The communication connection relationship of at least one network device among the network devices, the fourth network device includes the operation status information of the third network device;
  • the connection relationship prediction module is further configured to: predict the third network device and the fourth network device according to the first network information, the second network information, the third network information, and the fourth network device.
  • the sending module is further configured to: transfer the communication connection relationship between the third network device and at least one network device among the plurality of network devices to the user equipment.
  • FIG. 16 is a schematic structural diagram of the execution device provided by the embodiment of the present application. Tablets, laptops, smart wearable devices, monitoring data processing equipment or servers, etc., are not limited here.
  • the execution device 1600 includes: a receiver 1601, a transmitter 1602, a processor 1603, and a memory 1604 (the number of processors 1603 in the execution device 1600 may be one or more, and one processor is taken as an example in FIG. 16 ) , where the processor 1603 may include an application processor 16031 and a communication processor 16032 .
  • the receiver 1601 , the transmitter 1602 , the processor 1603 and the memory 1604 may be connected through a bus or in other ways.
  • the memory 1604 may include read-only memory and random-access memory, and provides instructions and data to the processor 1603 .
  • a part of the memory 1604 may also include a non-volatile random access memory (non-volatile random access memory, NVRAM).
  • NVRAM non-volatile random access memory
  • the memory 1604 stores processors and operating instructions, executable modules or data structures, or their subsets, or their extended sets, wherein the operating instructions may include various operating instructions for implementing various operations.
  • the processor 1603 controls the operations of the execution device.
  • various components of the execution device are coupled together through a bus system, where the bus system may include not only a data bus, but also a power bus, a control bus, and a status signal bus.
  • the various buses are referred to as bus systems in the figures.
  • the methods disclosed in the foregoing embodiments of the present application may be applied to the processor 1603 or implemented by the processor 1603 .
  • the processor 1603 may be an integrated circuit chip, which has a signal processing capability.
  • each step of the above method may be implemented by an integrated logic circuit of hardware in the processor 1603 or instructions in the form of software.
  • the above-mentioned processor 1603 can be a general-purpose processor, a digital signal processor (digital signal processing, DSP), a microprocessor or a microcontroller, and can further include an application-specific integrated circuit (application specific integrated circuit, ASIC), field programmable Field-programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.
  • DSP digital signal processing
  • ASIC application specific integrated circuit
  • FPGA field programmable Field-programmable gate array
  • the processor 1603 may implement or execute various methods, steps, and logic block diagrams disclosed in the embodiments of the present application.
  • a general-purpose processor may be a microprocessor, or the processor may be any conventional processor, or the like.
  • the steps of the method disclosed in connection with the embodiments of the present application may be directly implemented by a hardware decoding processor, or implemented by a combination of hardware and software modules in the decoding processor.
  • the software module can be located in a mature storage medium in the field such as random access memory, flash memory, read-only memory, programmable read-only memory or electrically erasable programmable memory, register.
  • the storage medium is located in the memory 1604, and the processor 1603 reads the information in the memory 1604, and completes the steps of the above method in combination with its hardware.
  • the receiver 1601 can be used to receive input digital or character information, and generate signal input related to performing device related settings and function control.
  • the transmitter 1602 can be used to output digital or character information through the first interface; the transmitter 1602 can also be used to send instructions to the disk group through the first interface to modify the data in the disk group; the transmitter 1602 can also include a display device such as a display screen .
  • the processor 1603 is configured to execute the above connection relationship prediction methods in FIG. 5 and FIG. 13 .
  • FIG. 17 is a schematic structural diagram of the training device provided in the embodiment of the present application.
  • the training device 1700 can be deployed with the image described in the embodiment corresponding to FIG. 17
  • the processing device is used to implement the functions of the data processing device in the embodiment corresponding to FIG. 18.
  • the training device 1700 is implemented by one or more servers.
  • the training device 1700 may have relatively large differences due to different configurations or performances, and may include One or more central processing units (central processing units, CPU) 1717 (for example, one or more processors) and memory 1732, one or more storage media 1730 for storing application programs 1742 or data 1744 (for example, one or more mass storage devices).
  • CPU central processing units
  • storage media 1730 for storing application programs 1742 or data 1744 (for example, one or more mass storage devices).
  • the memory 1732 and the storage medium 1730 may be temporary storage or persistent storage.
  • the program stored in the storage medium 1730 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations on the training device.
  • the central processing unit 1717 may be configured to communicate with the storage medium 1730 , and execute a series of instruction operations in the storage medium 1730 on the training device 1700 .
  • the training device 1700 can also include one or more power supplies 1726, one or more wired or wireless network interfaces 1750, one or more input and output interfaces 1758; or, one or more operating systems 1741, such as Windows ServerTM, Mac OS XTM , UnixTM, LinuxTM, FreeBSDTM and so on.
  • operating systems 1741 such as Windows ServerTM, Mac OS XTM , UnixTM, LinuxTM, FreeBSDTM and so on.
  • the central processing unit 1717 is configured to execute the above-mentioned model training method in FIG. 12 .
  • the embodiment of the present application also provides a computer program product, which, when running on a computer, causes the computer to perform the steps performed by the aforementioned execution device, or enables the computer to perform the steps performed by the aforementioned training device.
  • An embodiment of the present application also provides a computer-readable storage medium, the computer-readable storage medium stores a program for signal processing, and when it is run on a computer, the computer executes the steps performed by the aforementioned executing device , or, causing the computer to perform the steps performed by the aforementioned training device.
  • the execution device, training device or terminal device provided in the embodiment of the present application may specifically be a chip.
  • the chip includes: a processing unit and a communication unit.
  • the processing unit may be, for example, a processor, and the communication unit may be, for example, an input/output interface, pins or circuits etc.
  • the processing unit can execute the computer-executed instructions stored in the storage unit, so that the chips in the execution device execute the data processing methods described in the above embodiments, or make the chips in the training device execute the data processing methods described in the above embodiments.
  • the storage unit is a storage unit in the chip, such as a register, a cache, etc.
  • the storage unit may also be a storage unit located outside the chip in the wireless access device, such as only Read-only memory (ROM) or other types of static storage devices that can store static information and instructions, random access memory (random access memory, RAM), etc.
  • ROM Read-only memory
  • RAM random access memory
  • FIG. 18 is a schematic structural diagram of a chip provided by the embodiment of the present application.
  • the chip can be represented as a neural network processor NPU 1800, and the NPU 1800 is mounted on the main CPU (Host CPU) as a coprocessor. CPU), the tasks are assigned by the Host CPU.
  • the core part of the NPU is the operation circuit 1803, and the operation circuit 1803 is controlled by the controller 1804 to extract matrix data in the memory and perform multiplication operations.
  • the operation circuit 1803 includes multiple processing units (Process Engine, PE).
  • arithmetic circuit 1803 is a two-dimensional systolic array.
  • the arithmetic circuit 1803 may also be a one-dimensional systolic array or other electronic circuits capable of performing mathematical operations such as multiplication and addition.
  • arithmetic circuit 1803 is a general-purpose matrix processor.
  • the operation circuit fetches the data corresponding to the matrix B from the weight memory 1802, and caches it in each PE in the operation circuit.
  • the operation circuit takes the data of matrix A from the input memory 1801 and performs matrix operation with matrix B, and the obtained partial or final results of the matrix are stored in the accumulator (accumulator) 1808 .
  • the unified memory 1806 is used to store input data and output data.
  • the weight data directly accesses the controller (Direct Memory Access Controller, DMAC) 1805 through the storage unit, and the DMAC is transferred to the weight storage 1802.
  • Input data is also transferred to unified memory 1806 by DMAC.
  • DMAC Direct Memory Access Controller
  • the BIU is the Bus Interface Unit, that is, the bus interface unit 1810, which is used for the interaction between the AXI bus and the DMAC and the instruction fetch buffer (Instruction Fetch Buffer, IFB) 1809.
  • IFB Instruction Fetch Buffer
  • the bus interface unit 1810 (Bus Interface Unit, BIU for short), is used for the instruction fetch memory 1809 to obtain instructions from the external memory, and is also used for the storage unit access controller 1805 to obtain the original data of the input matrix A or the weight matrix B from the external memory.
  • BIU Bus Interface Unit
  • the DMAC is mainly used to move the input data in the external memory DDR to the unified memory 1806 , to move the weight data to the weight memory 1802 , or to move the input data to the input memory 1801 .
  • the vector calculation unit 1807 includes a plurality of calculation processing units, and further processes the output of the calculation circuit, such as vector multiplication, vector addition, exponential operation, logarithmic operation, size comparison, etc., if necessary. It is mainly used for non-convolutional/full-connection layer network calculations in neural networks, such as Batch Normalization (batch normalization), pixel-level summation, and feature balance Face upsampling, etc.
  • the vector computation unit 1807 can store the vector of the processed output to unified memory 1806 .
  • the vector calculation unit 1807 can apply a linear function; or, a nonlinear function to the output of the operation circuit 1803, such as performing linear interpolation on the feature plane extracted by the convolutional layer, and for example, a vector of accumulated values to generate an activation value.
  • the vector computation unit 1807 generates normalized values, pixel-level summed values, or both.
  • the vector of processed outputs can be used as an activation input to operational circuitry 1803, eg, for use in subsequent layers in a neural network.
  • An instruction fetch buffer (instruction fetch buffer) 1809 connected to the controller 1804 is used to store instructions used by the controller 1804;
  • the unified memory 1806, the input memory 1801, the weight memory 1802 and the fetch memory 1809 are all On-Chip memories. External memory is private to the NPU hardware architecture.
  • the processor mentioned above can be a general-purpose central processing unit, microprocessor, ASIC, or one or more integrated circuits for controlling the execution of the above-mentioned programs.
  • the device embodiments described above are only illustrative, and the units described as separate components may or may not be physically separated, and the components shown as units may or may not be A physical unit can be located in one place, or it can be distributed to multiple network units. Part or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.
  • the connection relationship between the modules indicates that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines.
  • the essence of the technical solution of this application or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product is stored in a readable storage medium, such as a floppy disk of a computer , U disk, mobile hard disk, ROM, RAM, magnetic disk or optical disk, etc., including several instructions to make a computer device (which can be a personal computer, training device, or network device, etc.) execute the instructions described in various embodiments of the present application method.
  • a computer device which can be a personal computer, training device, or network device, etc.
  • all or part of them may be implemented by software, hardware, firmware or any combination thereof.
  • software When implemented using software, it may be implemented in whole or in part in the form of a computer program product.
  • the computer program product includes one or more computer instructions.
  • the computer can be a general purpose computer, a special purpose computer, a computer network, or other programmable devices.
  • the computer instructions may be stored in or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions may be transferred from a website, computer, training device, or data The center communicates to the Another website site, computer, training device or data center for transmission.
  • the computer-readable storage medium may be any available medium that can be stored by a computer, or a data storage device such as a training device or a data center integrated with one or more available media.
  • the available medium may be a magnetic medium (for example, a floppy disk, a hard disk, or a magnetic tape), an optical medium (for example, DVD), or a semiconductor medium (for example, a solid state disk (Solid State Disk, SSD)) and the like.
  • a magnetic medium for example, a floppy disk, a hard disk, or a magnetic tape
  • an optical medium for example, DVD
  • a semiconductor medium for example, a solid state disk (Solid State Disk, SSD)

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Data Mining & Analysis (AREA)
  • General Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Data Exchanges In Wide-Area Networks (AREA)

Abstract

本申请涉及人工智能领域，公开了一种连接关系预测方法，方法包括：获取目标网络的第一网络信息以及第二网络信息，目标网络包括第一网络设备和第二网络设备，第一网络信息包括第一网络设备和第二网络设备所在的通信链路的拓扑结构信息，第二网络信息包括第一网络设备和第二网络设备在内的多个网络设备的运行状态信息；根据第一网络信息以及第二网络信息，预测第一网络设备和第二网络设备之间的通信连接关系。本申请可以增加连接关系的预测精度。

Description

一种连接关系预测方法及相关设备
本申请要求于2022年1月30日提交中国专利局、申请号为202210114738.5、发明名称为“一种连接关系预测方法及相关设备”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及人工智能领域，尤其涉及一种连接关系预测方法及相关设备。
背景技术
人工智能(artificial intelligence，AI)是利用数字计算机或者数字计算机控制的机器模拟、延伸和扩展人的智能，感知环境、获取知识并使用知识获得最佳结果的理论、方法、技术及应用系统。换句话说，人工智能是计算机科学的一个分支，它企图了解智能的实质，并生产出一种新的能以人类智能相似的方式作出反应的智能机器。人工智能也就是研究各种智能机器的设计原理与实现方法，使机器具有感知、推理与决策的功能。
网络拓扑是服务工程师最基本、最广泛的网络信息依据，准确的、完整的、实时度高的业务路径拓扑还原，在运维领域非常有竞争力。拓扑模块上叠加告警、性能等叠加信息，可以快速实现故障的诊断。拓扑也是网络可视化最基本的元素。但是目前的拓扑不完整以及不准确在各个局点和场景中普遍存在。以电信网络为例，电信网络可以包括核心网、数据承载网、传送网、无线接入网、固定接入网等网域，当前的电信网络能够生成单个网域的网络拓扑图，举例来说，单个网域内的网元是由该单个网域中的网管系统来配置，因此该单个网域内的网管系统可以根据配置网元时记录的配置信息生成该单个网域的网络拓扑信息，该拓扑信息用于描述单个网域内的网元之间的连接关系。
进一步地，如果要获得多个网域整体的网络拓扑图，则需要靠人工收集不同网域之间的节点连接关系和各个网域内的网络拓扑信息，并由人工根据收集到的节点连接关系以及各个域内的网络拓扑信息绘制多个网域整体的网络拓扑图，图1为对应的场景示意图。采用这种方式生成多个网域整体的网络拓扑图，效率低、正确率低，无法满足用户需求。
发明内容
本申请实施例提供了一种连接关系预测方法，将携带和第一网络设备以及第二网络设备相关的拓扑结构信息的第一网络信息，以及携带第一网络设备以及第二网络设备相关的运行状态信息的第二网络信息作为预测第一网络设备以及第二网络设备之间是否存在通信连接关系的参考，在预测网络设备之间的连接关系时，可以识别出更丰富的数据关联关系，进而增加连接关系的预测精度。
第一方面，本申请提供了一种连接关系预测方法，所述方法包括：
获取目标网络的第一网络信息以及第二网络信息；其中，所述目标网络包括多个网络设备，所述多个网络设备包括第一网络设备和第二网络设备，所述第一网络信息包括所述第一网络设备和所述第二网络设备所在的通信链路的拓扑结构信息，所述第二网络信息包 括所述第一网络设备和所述第二网络设备在内的多个网络设备的运行状态信息；
其中，第一网络设备和第二网络设备可以为目标网络中任意的两个网络设备。
其中，第一网络设备和第二网络设备可以为用户指定的。例如，用户可以指定预测目标网络中哪两个网络设备之间的通信连接关系。
在一种可能的实现中，可以获取到目标网络的第一网络信息，其中，第一网络信息可以包括所述目标网络中多个网络设备之间的通信连接关系。
其中，网络设备之间存在通信连接关系可以理解为：网络设备之间存在物理的通信路径(例如有线通信路径或者无线通信路径等)，或者是网络设备之间存在流量交互。
根据所述第一网络信息以及所述第二网络信息，通过特征提取，得到所述第一网络设备以及所述第二网络设备的目标特征向量；
根据所述目标特征向量，通过目标神经网络，预测所述第一网络设备和所述第二网络设备之间的通信连接关系。
从理论上来说，当两个网络设备之间的运行状态信息相似度较高时，则大概率认为两个网络设备之间存在通信连接关系的可能性越大，例如，当两个网络设备都有一个共同的网络设备，其三者之间的运行状态信息相似度都很大，则该两个网络设备之间的通信连接关系的概率较大，当然网络的结构很复杂，存在着比上述信息更丰富的判断规则，因此本申请可以将包括第一网络设备以及第二网络设备在内的多个网络设备的运行状态信息也作为判断第一网络设备以及第二网络设备是否存在通讯连接关系所考虑的因素。
从理论上来说，当两个网络设备之间的运行状态信息相似度较高时，则大概率认为两个网络设备之间存在通信连接关系的可能性越大，例如，当两个网络设备都有一个共同的网络设备，其三者之间的运行状态信息相似度都很大，则该两个网络设备之间的通信连接关系的概率较大，当然网络的结构很复杂，存在着比上述信息更丰富的判断规则，因此本申请可以将包括第一网络设备以及第二网络设备在内的多个网络设备的运行状态信息也作为判断第一网络设备以及第二网络设备是否存在通讯连接关系所考虑的因素。
本申请实施例提供了一种连接关系预测方法，所述方法包括：获取目标网络的第一网络信息以及第二网络信息；其中，所述目标网络包括多个网络设备，所述多个网络设备包括第一网络设备和第二网络设备，所述第一网络信息包括所述第一网络设备和所述第二网络设备所在的通信链路的拓扑结构信息，所述第二网络信息包括所述第一网络设备和所述第二网络设备在内的多个网络设备的运行状态信息；根据所述第一网络信息以及所述第二网络信息，通过特征提取，得到所述第一网络设备以及所述第二网络设备的目标特征向量；根据所述目标特征向量，通过目标神经网络，预测所述第一网络设备和所述第二网络设备之间的通信连接关系。本申请将携带和第一网络设备以及第二网络设备相关的拓扑结构信息的第一网络信息，以及携带第一网络设备以及第二网络设备相关的运行状态信息的第二网络信息作为预测第一网络设备以及第二网络设备之间是否存在通信连接关系的参考，在预测网络设备之间的连接关系时，可以识别出更丰富的数据关联关系，进而增加连接关系的预测精度。
在一种可能的实现中，所述告警信息包括如下信息的至少一种：网络设备发生告警的时间、网络设备发生的告警类型。
在一种可能的实现中，告警信息可以为网络设备的告警记录，目标网络包括多个设备，每条告警记录可以包含告警的名称、告警出现的时间和告警出现的网络设备的标识。
告警的名称又可以称为告警的类型；告警出现的设备的标识的形式可以有多种，本申请实施例对此不做具体限定，例如，告警出现的设备的标识可以是编号。
应理解，告警记录通常存在于告警日志中。所以，可以基于目标网络的告警日志获取告警记录。
作为一种实现方式，可以获取目标网络在目标时间段内产生的告警记录，其中目标时间段可以根据实际需要进行设定。例如，目标时间段可以为一个月，除此之外，目标时间段还可以为20天、25天、35天、40天等。
在一种可能的实现中，网络设备的关键性能指标可以为网络设备的通信流量。
在一种可能的实现中，所述根据所述第一网络信息以及所述第二网络信息，通过特征提取，得到所述第一网络设备以及所述第二网络设备的目标特征向量，包括：
根据所述第二网络信息，得到所述第一网络设备和所述第二网络设备在内的多个网络设备的多个嵌入向量；其中，运行状态信息的相似度越大的网络设备所对应的嵌入向量之间的距离越近；所述多个嵌入向量包括所述第一网络设备的第一嵌入向量、以及所述第二网络设备的第二嵌入向量；
根据所述第一网络信息、所述第一嵌入向量以及所述第二嵌入向量，通过特征提取，得到所述第一网络设备以及所述第二网络设备的目标特征向量。
在一种可能的实现中，可以将运行状态信息映射到嵌入向量的空间当中，也就是更高维的特征空间中，进而可以通过嵌入向量的形式来表达出运行状态信息，其中映射的目的可以为：具备更高相似度运行状态信息的网络设备之间对应的嵌入向量的相似度是更高的。
在一种可能的实现中，所述第一网络设备和所述第二网络设备在内的多个网络设备，包括：除所述第一网络设备和所述第二网络设备之外的至少一个网络设备。例如，所述第一网络设备和所述第二网络设备在内的多个网络设备，可以包括目标网络所包括的全部网络设备。
在一种可能的实现中，所述根据所述第二网络信息，得到所述第一网络设备和所述第二网络设备在内的多个网络设备的多个嵌入向量，包括：
根据所述第二网络信息，得到所述第一网络设备和所述第二网络设备在内的多个网络设备的语义拓扑，所述语义拓扑包括网络设备之间的语义连接关系，其中，所述运行状态信息的相似度大于阈值的网络设备之间存在所述语义连接关系；
对所述语义拓扑进行随机游走，以得到所述第一网络设备和所述第二网络设备在内的多个网络设备的多个嵌入向量。
在一种可能的实现中，所述根据所述第一网络信息、所述第一嵌入向量以及所述第二嵌入向量，通过特征提取，得到所述第一网络设备以及所述第二网络设备的目标特征向量，包括：根据所述第一网络信息、所述语义拓扑的信息、所述第一嵌入向量以及所述第二嵌入向量，通过特征提取，得到所述第一网络设备以及所述第二网络设备的目标特征向量。
在一种可能的实现中，所述获取目标网络的第一网络信息以及第二网络信息，包括：获取来自用户设备的所述目标网络的第一网络信息以及第二网络信息；
所述方法还包括：在所述预测所述第一网络设备和所述第二网络设备之间的通信连接关系之后，将所述第一网络设备和所述第二网络设备之间的通信连接关系传递至所述用户设备。
在一种可能的实现中，所述特征提取为基于特征提取网络实现的，所述特征提取网络为图神经网络GNN；所述目标神经网络为全连接网络。
第二方面，本申请提供了一种连接关系预测方法，所述方法包括：
接收来自用户设备的目标网络的第一网络信息以及第二网络信息；其中，所述目标网络包括多个网络设备，所述第一网络信息包括所述多个网络设备之间通信连接的拓扑结构信息，所述第二网络信息包括所述多个网络设备的运行状态信息；所述多个网络设备包括第一网络设备和第二网络设备；
根据所述第一网络信息以及所述第二网络信息，预测所述第一网络设备和所述第二网络设备之间的通信连接关系；
将所述所述第一网络设备和所述第二网络设备之间的通信连接关系传递至所述用户设备。
在一种可能的实现中，所述运行状态信息包括如下信息的至少一种：
网络设备的告警信息、网络设备的关键性能指标KPI。
在一种可能的实现中，所述告警信息包括如下信息的至少一种：
网络设备发生告警的时间、网络设备发生的告警类型。
在一种可能的实现中，所述方法还包括：
接收来自用户设备的第三网络设备的第三网络信息和第四网络信息，所述第三网络设备不属于所述目标网络，所述第三网络信息包括所述第三网络信息与所述多个网络设备中至少一个网络设备的通信连接关系，所述第四网络设备包括所述第三网络设备的运行状态信息；
根据所述第一网络信息、所述第二网络信息、所述第三网络信息和所述第四网络设备， 预测所述第三网络设备和所述多个网络设备中至少一个网络设备之间的通信连接关系；
将所述第三网络设备和所述多个网络设备中至少一个网络设备之间的通信连接关系传递至所述用户设备。
第三方面，本申请提供了一种连接关系预测装置，所述装置包括：
获取模块，用于获取目标网络的第一网络信息以及第二网络信息；其中，所述目标网络包括多个网络设备，所述多个网络设备包括第一网络设备和第二网络设备，所述第一网络信息包括所述第一网络设备和所述第二网络设备所在的通信链路的拓扑结构信息，所述第二网络信息包括所述第一网络设备和所述第二网络设备在内的多个网络设备的运行状态信息；
特征提取模块，用于根据所述第一网络信息以及所述第二网络信息，通过特征提取，得到所述第一网络设备以及所述第二网络设备的目标特征向量；
连接关系预测模块，用于根据所述目标特征向量，通过目标神经网络，预测所述第一网络设备和所述第二网络设备之间的通信连接关系。
在一种可能的实现中，所述运行状态信息包括如下信息的至少一种：
网络设备的告警信息、网络设备的关键性能指标KPI。
在一种可能的实现中，所述告警信息包括如下信息的至少一种：
网络设备发生告警的时间、网络设备发生的告警类型。
在一种可能的实现中，所述特征提取模块，具体用于：
根据所述第二网络信息，得到所述第一网络设备和所述第二网络设备在内的多个网络设备的多个嵌入向量；其中，运行状态信息的相似度越大的网络设备所对应的嵌入向量之间的距离越近；所述多个嵌入向量包括所述第一网络设备的第一嵌入向量、以及所述第二网络设备的第二嵌入向量；
根据所述第一网络信息、所述第一嵌入向量以及所述第二嵌入向量，通过特征提取，得到所述第一网络设备以及所述第二网络设备的目标特征向量。
在一种可能的实现中，所述第一网络设备和所述第二网络设备在内的多个网络设备，包括：
除所述第一网络设备和所述第二网络设备之外的至少一个网络设备。
在一种可能的实现中，所述特征提取模块，具体用于：
根据所述第二网络信息，得到所述第一网络设备和所述第二网络设备在内的多个网络设备的语义拓扑，所述语义拓扑包括网络设备之间的语义连接关系，其中，所述运行状态信息的相似度大于阈值的网络设备之间存在所述语义连接关系；
对所述语义拓扑进行随机游走，以得到所述第一网络设备和所述第二网络设备在内的多个网络设备的多个嵌入向量。
在一种可能的实现中，所述特征提取模块，具体用于：
根据所述第一网络信息、所述语义拓扑的信息、所述第一嵌入向量以及所述第二嵌入向量，通过特征提取，得到所述第一网络设备以及所述第二网络设备的目标特征向量。
在一种可能的实现中，所述获取模块，具体用于：
获取来自用户设备的所述目标网络的第一网络信息以及第二网络信息；
所述装置还包括：
发送模块，用于在所述预测所述第一网络设备和所述第二网络设备之间的通信连接关系之后，将所述第一网络设备和所述第二网络设备之间的通信连接关系传递至所述用户设备。
在一种可能的实现中，所述特征提取为基于特征提取网络实现的，所述特征提取网络为图神经网络GNN；所述目标神经网络为全连接网络。
第四方面，本申请提供了一种连接关系预测装置，所述装置包括：
获取模块，用于接收来自用户设备的目标网络的第一网络信息以及第二网络信息；其中，所述目标网络包括多个网络设备，所述第一网络信息包括所述多个网络设备之间通信连接的拓扑结构信息，所述第二网络信息包括所述多个网络设备的运行状态信息；所述多个网络设备包括第一网络设备和第二网络设备；
连接关系预测模块，用于根据所述第一网络信息以及所述第二网络信息，预测所述第一网络设备和所述第二网络设备之间的通信连接关系；
发送模块，用于将所述所述第一网络设备和所述第二网络设备之间的通信连接关系传递至所述用户设备。
在一种可能的实现中，所述运行状态信息包括如下信息的至少一种：
网络设备的告警信息、网络设备的关键性能指标KPI。
在一种可能的实现中，所述告警信息包括如下信息的至少一种：
网络设备发生告警的时间、网络设备发生的告警类型。
在一种可能的实现中，所述获取模块，还用于：
接收来自用户设备的第三网络设备的第三网络信息和第四网络信息，所述第三网络设备不属于所述目标网络，所述第三网络信息包括所述第三网络信息与所述多个网络设备中至少一个网络设备的通信连接关系，所述第四网络设备包括所述第三网络设备的运行状态 信息；
所述连接关系预测模块，还用于：根据所述第一网络信息、所述第二网络信息、所述第三网络信息和所述第四网络设备，预测所述第三网络设备和所述多个网络设备中至少一个网络设备之间的通信连接关系；
所述发送模块，还用于：将所述第三网络设备和所述多个网络设备中至少一个网络设备之间的通信连接关系传递至所述用户设备。
第五方面，本申请实施例提供了一种连接关系预测装置，可以包括存储器、处理器以及总线系统，其中，存储器用于存储程序，处理器用于执行存储器中的程序，以执行如上述第一方面及其任一可选的方法、第二方面及其任一可选的方法。
第六方面，本申请实施例提供了一种计算机可读存储介质，所述计算机可读存储介质中存储有计算机程序，当其在计算机上运行时，使得计算机执行上述第一方面及其任一可选的方法、第二方面及其任一可选的方法。
第七方面，本申请实施例提供了一种计算机程序，当其在计算机上运行时，使得计算机执行上述第一方面及其任一可选的方法、第二方面及其任一可选的方法。
第八方面，本申请提供了一种芯片系统，该芯片系统包括处理器，用于支持模型蒸馏装置实现上述方面中所涉及的功能，例如，发送或处理上述方法中所涉及的数据；或，信息。在一种可能的设计中，所述芯片系统还包括存储器，所述存储器，用于保存执行设备或训练设备必要的程序指令和数据。该芯片系统，可以由芯片构成，也可以包括芯片和其他分立器件。
本申请实施例提供了一种连接关系预测方法，所述方法包括：获取目标网络的第一网络信息以及第二网络信息；其中，所述目标网络包括多个网络设备，所述多个网络设备包括第一网络设备和第二网络设备，所述第一网络信息包括所述第一网络设备和所述第二网络设备所在的通信链路的拓扑结构信息，所述第二网络信息包括所述第一网络设备和所述第二网络设备在内的多个网络设备的运行状态信息；根据所述第一网络信息以及所述第二网络信息，通过特征提取，得到所述第一网络设备以及所述第二网络设备的目标特征向量；根据所述目标特征向量，通过目标神经网络，预测所述第一网络设备和所述第二网络设备之间的通信连接关系。本申请将携带和第一网络设备以及第二网络设备相关的拓扑结构信息的第一网络信息，以及携带第一网络设备以及第二网络设备相关的运行状态信息的第二网络信息作为预测第一网络设备以及第二网络设备之间是否存在通信连接关系的参考，在预测网络设备之间的连接关系时，可以识别出更丰富的数据关联关系，进而增加连接关系的预测精度。
附图说明
图1为人工智能主体框架的一种结构示意图；
图2为一种网络结构的示意图；
图3为一种网络结构的示意图；
图4为一种应用架构的示意图；
图5为一种连接关系预测方法的流程示意图；
图6为一种第一网络信息的示意；
图7为一种节点的标注示意；
图8为一个对第一网络信息进行编码的流程示意；
图9为一个语义拓扑的构建示意；
图10为一个基于随机游走来构建嵌入向量的示意；
图11为一个特征融合的示意；
图12为本申请实施例提供的一种模型训练的流程示意；
图13为本申请实施例提供的一种连接关系预测方法的流程示意；
图14为本申请实施例提供的一种连接关系预测装置的结构示意；
图15为本申请实施例提供的一种连接关系预测装置的结构示意；
图16为本申请实施例提供的执行设备的一种结构示意图；
图17是本申请实施例提供的训练设备一种结构示意图；
图18为本申请实施例提供的芯片的一种结构示意图。
具体实施方式
下面结合本发明实施例中的附图对本发明实施例进行描述。本发明的实施方式部分使用的术语仅用于对本发明的具体实施例进行解释，而非旨在限定本发明。
下面结合附图，对本申请的实施例进行描述。本领域普通技术人员可知，随着技术的发展和新场景的出现，本申请实施例提供的技术方案对于类似的技术问题，同样适用。
本申请的说明书和权利要求书及上述附图中的术语“第一”、“第二”等是用于区别类似的对象，而不必用于描述特定的顺序或先后次序。应该理解这样使用的术语在适当情况下可以互换，这仅仅是描述本申请的实施例中对相同属性的对象在描述时所采用的区分方式。此外，术语“包括”和“具有”以及他们的任何变形，意图在于覆盖不排他的包含，以便包含一系列单元的过程、方法、系统、产品或设备不必限于那些单元，而是可包括没有清楚地列出的或对于这些过程、方法、产品或设备固有的其它单元。
首先对人工智能系统总体工作流程进行描述，请参见图2，图2示出的为人工智能主体框架的一种结构示意图，下面从“智能信息链”(水平轴)和“IT价值链”(垂直轴)两个维度对上述人工智能主题框架进行阐述。其中，“智能信息链”反映从数据的获取到处理的一列过程。举例来说，可以是智能信息感知、智能信息表示与形成、智能推理、智能决策、智能执行与输出的一般过程。在这个过程中，数据经历了“数据—信息—知识—智慧”的凝练过程。“IT价值链”从人智能的底层基础设施、信息(提供和处理技术实现)到系统的产业生态过程，反映人工智能为信息技术产业带来的价值。
(1)基础设施
基础设施为人工智能系统提供计算能力支持，实现与外部世界的沟通，并通过基础平台实现支撑。通过传感器与外部沟通；计算能力由智能芯片(CPU、NPU、GPU、ASIC、 FPGA等硬件加速芯片)提供；基础平台包括分布式计算框架及网络等相关的平台保障和支持，可以包括云存储和计算、互联互通网络等。举例来说，传感器和外部沟通获取数据，这些数据提供给基础平台提供的分布式计算系统中的智能芯片进行计算。
(2)数据
基础设施的上一层的数据用于表示人工智能领域的数据来源。数据涉及到图形、图像、语音、文本，还涉及到传统设备的物联网数据，包括已有系统的业务数据以及力、位移、液位、温度、湿度等感知数据。
(3)数据处理
数据处理通常包括数据训练，机器学习，深度学习，搜索，推理，决策等方式。
其中，机器学习和深度学习可以对数据进行符号化和形式化的智能信息建模、抽取、预处理、训练等。
推理是指在计算机或智能系统中，模拟人类的智能推理方式，依据推理控制策略，利用形式化的信息进行机器思维和求解问题的过程，典型的功能是搜索与匹配。
决策是指智能信息经过推理后进行决策的过程，通常提供分类、排序、预测等功能。
(4)通用能力
对数据经过上面提到的数据处理后，进一步基于数据处理的结果可以形成一些通用的能力，比如可以是算法或者一个通用系统，例如，翻译，文本的分析，计算机视觉的处理，语音识别，图像的识别等等。
(5)智能产品及行业应用
智能产品及行业应用指人工智能系统在各领域的产品和应用，是对人工智能整体解决方案的封装，将智能信息决策产品化、实现落地应用，其应用领域主要包括：智能终端、智能交通、智能医疗、自动驾驶、智慧城市等。
接下来介绍本申请实施例的应用场景：
请参见图3，图3是本申请实施例提供的一种网络系统(例如本申请实施例中的目标网络)的架构示意图，该网络系统可以包括第一节点201、第二节点202以及多个网域203，该多个网域203可以包括核心网、数据承载网、传送网、无线接入网、固定接入网中的两个或两个以上网域，例如，包括其中2个网域、或者3个网域、或者4个网域、或者5个网域，等等。每个网域203中包括多个网元204(本申请实施例中的网元也可以称之为网络设备，例如第一网络设备、第二网络设备以及第三网络设备)，该多个网元204具体包括多少个网元此处不作限定；图2中仅对一个网域内的网元进行了示意，其余网域内的结构可以依此类推。另外，不同的网域203包括的网元204的数量可以相同也可以不同，不同网域203包括的网元204的类型通常不同(但也可能出现相同类型网元的情况)，举例来说，核心网可以包括分组交换(packet switched，PS)、电路交换(circuit switched，CS)、归属用户服务器(home subscriber server，HSS)等系列类型网元，数据承载网可以包括交换机、路由器、防火墙等系列类型网元、传送网可以包括微波、多业务传送平台(multi-service transmission platform，MSTP)、波分、分组传送网(packet transport network，PTN)等系列类型网元，无线接入网可以包括第二代移动通信技术(The 2nd-Generation，2G)、第三代移动 通信技术(The 3rd-Genera tion，3G)、第四代移动通信技术(the 4th Generation mobile communication，4G)、第五代移动通信技术(the 5th-Generation，5G)中的基站、及基站控制器等系列类型网元，固定接入网可以包括光线路终端(optical line terminal，OLT)、光网络终端(optical network terminal，ONT)、多住户单元(multiple x unit，MxU)等系列类型网元，等等。
其中，第一节点201可以为终端设备、服务器等用户设备，该第一节点201可以获取到网络系统的网络拓扑以及运行状态信息，并可以将网络拓扑(指示网元之间的通信连接关系)以及运行状态信息传递至第二节点202，第二节点202可以为服务器或者终端设备，该第二节点202可以为第一节点201提供网络系统中网元之间连接关系的正误判断以及补全的服务，也就是说，第二节点202可以根据第一节点201上传的信息，来判断第一节点101上传的网络拓扑所指示的网元之间的通信连接关系是否准确或者是否存在遗漏，并将判断结果传递至第一节点101。
例如，第二节点202可以通过应用程序接口(application programming interface，API)来为第一节点201提供上述连接关系的预测服务。
例如，第二节点202可以为终端设备或者具备云计算能力的设备，第二节点202上可以安装有用于提供上述连接关系的预测服务的应用程序，通过和第一节点201之间的交互来为第一节点201提供上述连接关系的预测服务。
由于本申请实施例涉及大量神经网络的应用，为了便于理解，下面先对本申请实施例涉及的相关术语及神经网络等相关概念进行介绍。
(1)神经网络
神经网络可以是由神经单元组成的，神经单元可以是指以xs(即输入数据)和截距1为输入的运算单元，该运算单元的输出可以为：
其中，s＝1、2、……n，n为大于1的自然数，Ws为xs的权重，b为神经单元的偏置。f为神经单元的激活函数(activation functions)，用于将非线性特性引入神经网络中，来将神经单元中的输入信号转换为输出信号。该激活函数的输出信号可以作为下一层卷积层的输入，激活函数可以是sigmoid函数。神经网络是将多个上述单一的神经单元联结在一起形成的网络，即一个神经单元的输出可以是另一个神经单元的输入。每个神经单元的输入可以与前一层的局部接受域相连，来提取局部接受域的特征，局部接受域可以是由若干个神经单元组成的区域。
(2)深度神经网络
深度神经网络(Deep Neural Network，DNN)，也称多层神经网络，可以理解为具有很多层隐含层的神经网络，这里的“很多”并没有特别的度量标准。从DNN按不同层的位置划分，DNN内部的神经网络可以分为三类：输入层，隐含层，输出层。一般来说第一层是输入层，最后一层是输出层，中间的层数都是隐含层。层与层之间是全连接的，也就是说，第i层的任意一个神经元一定与第i+1层的任意一个神经元相连。虽然DNN看起来很复杂， 但是就每一层的工作来说，其实并不复杂，简单来说就是如下线性关系表达式：其中，是输入向量，是输出向量，是偏移向量，W是权重矩阵(也称系数)，α()是激活函数。每一层仅仅是对输入向量经过如此简单的操作得到输出向量由于DNN层数多，则系数W和偏移向量的数量也就很多了。这些参数在DNN中的定义如下所述：以系数W为例：假设在一个三层的DNN中，第二层的第4个神经元到第三层的第2个神经元的线性系数定义为上标3代表系数W所在的层数，而下标对应的是输出的第三层索引2和输入的第二层索引4。总结就是：第L-1层的第k个神经元到第L层的第j个神经元的系数定义为需要注意的是，输入层是没有W参数的。在深度神经网络中，更多的隐含层让网络更能够刻画现实世界中的复杂情形。理论上而言，参数越多的模型复杂度越高，“容量”也就越大，也就意味着它能完成更复杂的学习任务。训练深度神经网络的也就是学习权重矩阵的过程，其最终目的是得到训练好的深度神经网络的所有层的权重矩阵(由很多层的向量W形成的权重矩阵)。
(3)图(Graph)：
图为包括至少一个顶点以及至少一条边的数据结构。在一些场景中，图中的顶点可以映射为实体，图中的边可以映射为实体与实体之间的关系。图可以是有向图或无向图。当然，图还可以包括顶点以及边以外的其他数据，例如顶点的标签以及边的标签等。在一个示例性场景中，应用于好友推荐的场景中，图中的每个顶点可以表示一个用户，图中的每条边可以表示不同用户之间的社交关系，图中每个顶点的数据为用户的画像数据以及用户的行为数据，例如用户的年龄、职业、爱好、学历等。又如，应用于在商品推荐的场景中，图中的每个顶点可以表示一个用户或一个商品，图中的每条边可以表示用户与商品之间的交互关系，例如购买关系、收藏关系等。又如，应用于金融风控的场景中，图中的每个顶点可以表示账号、交易或资金。图中的边可以表示资金的流动关系，例如图中的环路可以表示循环转账。再如，应用于网络系统中网元之间连接关系确定的场景中，图中的每个顶点可以表示一个网元，例如路由器、交换机、终端等，图中的每条边可以表示不同网元之间的连接关系。
(4)子图(英文：subgraph)：
为图的一部分，包括图中的部分顶点以及部分边。子图也可以称为图中的分区(英文：partition)。一个图可以包括多个子图。
(5)图神经网络(graph neural network，GNN)：
GNN是一种带有结构信息的深度学习方法，可以用于计算节点当前的状态。图神经网络的信息传递按照给定的图结构进行，可以根据相邻节点更新每个节点的状态。具体地，其可以根据当前节点的结构图，以神经网络作为点信息的聚合函数，将所有相邻节点的信息传递到当前节点，结合当前节点的状态进行更新。图神经网络的输出为所有节点的状态。
(6)损失函数
在训练深度神经网络的过程中，因为希望深度神经网络的输出尽可能的接近真正想要预测的值，所以可以通过比较当前网络的预测值和真正想要的目标值，再根据两者之间的 差异情况来更新每一层神经网络的权重向量(当然，在第一次更新之前通常会有初始化的过程，即为深度神经网络中的各层预先配置参数)，比如，如果网络的预测值高了，就调整权重向量让它预测低一些，不断的调整，直到深度神经网络能够预测出真正想要的目标值或与真正想要的目标值非常接近的值。因此，就需要预先定义“如何比较预测值和目标值之间的差异”，这便是损失函数(loss function)或目标函数(objective function)，它们是用于衡量预测值和目标值的差异的重要方程。其中，以损失函数举例，损失函数的输出值(loss)越高表示差异越大，那么深度神经网络的训练就变成了尽可能缩小这个loss的过程。
(7)反向传播算法
卷积神经网络可以采用误差反向传播(back propagation，BP)算法在训练过程中修正初始的超分辨率模型中参数的大小，使得超分辨率模型的重建误差损失越来越小。具体地，前向传递输入信号直至输出会产生误差损失，通过反向传播误差损失信息来更新初始的超分辨率模型中参数，从而使误差损失收敛。反向传播算法是以误差损失为主导的反向传播运动，旨在得到最优的超分辨率模型的参数，例如权重矩阵。
接下来介绍本申请实施例中执行连接关系预测方法的执行主体的更细节的架构。
下面结合图4对本申请实施例提供的系统架构进行详细的介绍。图4为本申请实施例提供的系统架构示意图。如图4所示，系统架构500包括执行设备510、训练设备520、数据库530、客户设备540、数据存储系统550以及数据采集系统560。
执行设备510包括计算模块511、I/O接口512、预处理模块513和预处理模块514。计算模块511中可以包括目标模型/规则501，预处理模块513和预处理模块514是可选的。
数据采集设备560用于采集训练样本。训练样本可以为图像数据、文本数据、音频数据等等，在本申请实施例中，训练样本为网络(例如目标网络)的拓扑结构信息以及网元的运行状态信息。在采集到训练样本之后，数据采集设备560将这些训练样本存入数据库530。
训练设备520可以基于数据库530中维护训练样本，对待训练的神经网络(例如本申请实施例中的特征提取网络、目标神经网格等)，以得到目标模型/规则501。
需要说明的是，在实际应用中，数据库530中维护的训练样本不一定都来自于数据采集设备560的采集，也有可能是从其他设备接收得到的。另外需要说明的是，训练设备520也不一定完全基于数据库530维护的训练样本进行目标模型/规则501的训练，也有可能从云端或其他地方获取训练样本进行模型训练，上述描述不应该作为对本申请实施例的限定。
根据训练设备520训练得到的目标模型/规则501可以应用于不同的系统或设备中，如应用于图4所示的执行设备510，该执行设备510可以是终端，如手机终端，平板电脑，笔记本电脑，增强现实(augmented reality，AR)/虚拟现实(virtual reality，VR)设备，车载终端等，还可以是服务器或者云端等。
具体的，训练设备520可以将训练后的模型传递至执行设备510。
在图4中，执行设备510配置输入/输出(input/output，I/O)接口512，用于与外部设备进行数据交互，用户可以通过客户设备540向I/O接口512输入数据(例如本申请实施例中的目标网络的第一网络信息、第二网络信息、第三网络信息以及第四网络信息)。
预处理模块513和预处理模块514用于根据I/O接口512接收到的输入数据进行预处理。应理解，可以没有预处理模块513和预处理模块514或者只有的一个预处理模块。当不存在预处理模块513和预处理模块514时，可以直接采用计算模块511对输入数据进行处理。
在执行设备510对输入数据进行预处理，或者在执行设备510的计算模块511执行计算等相关的处理过程中，执行设备510可以调用数据存储系统550中的数据、代码等以用于相应的处理，也可以将相应处理得到的数据、指令等存入数据存储系统550中。
最后，I/O接口512将处理结果(例如本申请实施例中的网络设备之间的连接关系)提供给客户设备540，从而提供给用户。
在图4所示情况下，用户可以手动给定输入数据，该“手动给定输入数据”可以通过I/O接口512提供的界面进行操作。另一种情况下，客户设备540可以自动地向I/O接口512发送输入数据，如果要求客户设备540自动发送输入数据需要获得用户的授权，则用户可以在客户设备540中设置相应权限。用户可以在客户设备540查看执行设备510输出的结果，具体的呈现形式可以是显示、声音、动作等具体方式。客户设备540也可以作为数据采集端，采集如图所示输入I/O接口512的输入数据及输出I/O接口512的输出结果作为新的样本数据，并存入数据库530。当然，也可以不经过客户设备540进行采集，而是由I/O接口512直接将如图所示输入I/O接口512的输入数据及输出I/O接口512的输出结果，作为新的样本数据存入数据库530。
值得注意的是，图4仅是本申请实施例提供的一种系统架构的示意图，图中所示设备、器件、模块等之间的位置关系不构成任何限制，例如，在图4中，数据存储系统550相对执行设备510是外部存储器，在其它情况下，也可以将数据存储系统550置于执行设备510中。应理解，上述执行设备510可以部署于客户设备540中。
从模型的推理侧来说：
本申请实施例中，上述执行设备520的计算模块511可以获取到数据存储系统550中存储的代码来实现本申请实施例中的连接关系预测方法。
本申请实施例中，执行设备520的计算模块511可以包括硬件电路(如专用集成电路(application specific integrated circuit，ASIC)、现场可编程门阵列(field-programmable gate array，FPGA)、通用处理器、数字信号处理器(digital signal processing，DSP)、微处理器或微控制器等等)、或这些硬件电路的组合，例如，训练设备520可以为具有执行指令功能的硬件系统，如CPU、DSP等，或者为不具有执行指令功能的硬件系统，如ASIC、FPGA等，或者为上述不具有执行指令功能的硬件系统以及具有执行指令功能的硬件系统的组合。
具体的，执行设备520的计算模块511可以为具有执行指令功能的硬件系统，本申请实施例提供的连接关系预测方法可以为存储在存储器中的软件代码，执行设备520的计算模块511可以从存储器中获取到软件代码，并执行获取到的软件代码来实现本申请实施例 提供的连接关系预测方法。
应理解，执行设备520的计算模块511可以为不具有执行指令功能的硬件系统以及具有执行指令功能的硬件系统的组合，本申请实施例提供的连接关系预测方法的部分步骤还可以通过执行设备520的计算模块511中不具有执行指令功能的硬件系统来实现，这里并不限定。
从模型的训练侧来说：
本申请实施例中，上述训练设备520可以获取到存储器(图4中未示出，可以集成于训练设备520或者与训练设备520分离部署)中存储的代码来实现本申请实施例中的连接关系预测方法。
本申请实施例中，训练设备520可以包括硬件电路(如专用集成电路(application specific integrated circuit，ASIC)、现场可编程门阵列(field-programmable gate array，FPGA)、通用处理器、数字信号处理器(digital signal processing，DSP)、微处理器或微控制器等等)、或这些硬件电路的组合，例如，训练设备520可以为具有执行指令功能的硬件系统，如CPU、DSP等，或者为不具有执行指令功能的硬件系统，如ASIC、FPGA等，或者为上述不具有执行指令功能的硬件系统以及具有执行指令功能的硬件系统的组合。
具体的，训练设备520可以为具有执行指令功能的硬件系统，本申请实施例提供的连接关系预测方法可以为存储在存储器中的软件代码，训练设备520可以从存储器中获取到软件代码，并执行获取到的软件代码来实现本申请实施例提供的连接关系预测方法。
应理解，训练设备520可以为不具有执行指令功能的硬件系统以及具有执行指令功能的硬件系统的组合，本申请实施例提供的中连接关系预测方法的部分步骤还可以通过训练设备520中不具有执行指令功能的硬件系统来实现，这里并不限定。
应理解，上述训练设备的数量可以为多个(每个作为一个计算节点)。
接下来从模型的推理侧结合附图介绍本申请实施例提供的一种连接关系预测方法。
参照图5，图5为本申请实施例提供的一种连接关系预测方法的流程示意，如图5所示，本申请实施例提供的连接关系预测方法包括：
501、获取目标网络的第一网络信息以及第二网络信息；其中，所述目标网络包括多个网络设备，所述多个网络设备包括第一网络设备和第二网络设备，所述第一网络信息包括所述第一网络设备和所述第二网络设备所在的通信链路的拓扑结构信息，所述第二网络信息包括所述第一网络设备和所述第二网络设备在内的多个网络设备的运行状态信息。
在一种可能的场景中，用户设备可以根据配置网元时记录的配置信息生成目标网络的拓扑信息(该拓扑信息可以指示目标网络中包括的多个网络设备以及网络设备之间的通信连接关系)，然而，由于拓扑信息是人工收集的，会存在连接关系不准确或者连接关系遗漏的情况。因此，可以将目标网络的相关数据传递至提供连接关系预测服务的设备(例如步骤501的执行主体)。
在一种可能的实现中，目标网络可以是任意一个通信网络。本申请实施例对目标网络的规模和目标网络的拓扑结构不做具体限定。
在一种可能的实现中，步骤501的执行主体可以为终端设备或者服务器。
在一种可能的实现中，服务器可以接收到来自用户发送的目标网络的第一网络信息以及第二网络信息，例如，服务器可以通过API或者云服务器提供的其他类型的接口来接收到用户的终端设备(或者服务器)发送的目标网络的第一网络信息以及第二网络信息，或者是服务器上安装的云客户端可以接收到用户的终端设备(或者服务器)发送的目标网络的第一网络信息以及第二网络信息。
在一种可能的实现中，终端设备上安装的应用程序的客户端可以接收到用户输入的目标网络的第一网络信息以及第二网络信息。
在一种可能的实现中，所述目标网络包括多个网络设备(或者称之为网元)。第一网络信息可以包括多个网络设备之间的通信连接关系。
在一种可能的实现中，可以获取到目标网络的第一网络信息，其中，第一网络信息可以包括所述目标网络中多个网络设备之间的通信连接关系。
其中，网络设备之间存在通信连接关系可以理解为：网络设备之间存在物理的通信路径(例如有线通信路径或者无线通信路径等)，或者是网络设备之间存在流量交互。
该通信连接关系包括直接通信连接和间接通信连接。直接通信连接可以理解为两个设备可以直接进行通信，间接通信连接可以理解为两个设备需要通过至少一个中间设备通信连接。例如，基站与微波设备之间的通信连接关系为直接通信连接；而基站与路由设备之间通过微波设备通信连接，所以基站与路由设备之间的通信连接关系为间接连接；同样地，基站与核心网设备之间通过微波设备、路由设备通信连接，所以基站与核心网设备之间的通信连接关系也为间接连接。
除了上述用户上传的方式之外，作为一种实现方式，基于网络中心的本地存储器通常用于存储目标网络的信息，进而可以从网络中心的本地存储器中获取目标网络的拓扑数据(第一网络信息)。
作为另一种实现方式，可以根据目标网络的设备路径日志获取目标网络的拓扑数据，设备路径日志中包含至少一条通信路径的数据，至少一条通信路径包含多个设备，每条通信路径包含多个通信连接的设备，每条通信路径的数据包含多个通信连接的设备的标识以及多个通信连接的设备的类型信息。
如下表一所示，通信路径的数据可以包括通信路径的编号、通信路径上的设备的名称、通信路径上的设备的类型以及设备在通信路径上的编号。
表一
在上表中，设备在通信路径上的编号可以具体表示通信路径上的任一设备到某一特定设备需要几跳。例如，在编号为1的通信路径上，特定设备为设备11，相应地，设备11在该通信路径上的编号为0；设备22与设备11连接，所以设备22在该通信路径上的编号为1，表示设备22到设备11需要一跳；设备33通过设备22与设备11连接，所以设备33在该通信路径上的编号为2，表示设备22到设备11需要两跳。
其中，通信路径的编号、通信路径上的设备的名称、通信路径上的设备的类型以及设备在通信路径上的编号不限于上表所示的形式。
以上表所示的路径日志为例，编号为1的通信路径上包含设备11、设备22和设备33，因此可以确定设备11、设备22和设备33之间具有通信连接关系；编号为2的通信路径上包含设备11和设备44，因此可以确定设备11还与设备44具有通信连接关系。由此可见，根据设备路径日志除了可以确定目标网络中多个设备的类型外，还可以确定目标网络中多个设备的通信连接关系。
在一种可能的实现中，第一网络信息可以包括多个通信链路的拓扑结构信息(具体可以表现为每个通信链路上的网络设备以及网络设备之间的连接顺序)，每个通信链路上的网络设备之间可以存在串行的连接关系。
其中，上述网络设备之间的连接顺序也可以描述为连接位置。例如，一条通信链路上可以包括网络设备A、网络设备B以及网络设备C，网络设备A和网络设备B之间存在通信连接关系，则网络设备B相对于网络设备A的连接位置为：不存在其他网络设备的直接连接，而网络设备C和网络设备A之间间隔着网络设备B，则网络设备C可以认为相对于网络设备A的连接位置为：间隔了一个网络设备的连接位置。
在一种可能的实现中，第一网络信息可以为上述通信链路的拓扑结构信息的形式，也可以为其他的表达形式，例如通过表述网络设备之间哪些网络设备存在通信连接关系，应当理解，上述信息，同样也可以蕴含有通信链路的信息。
例如仍然以上述的通信链路为例，第一网络信息可以包括：第一网络设备A和网络设备B之间存在通信连接关系，网络设备B和网络设备C之间存在通讯连接关系，上述信息可以隐含得知网络设备A、网络设备B和网络设备C在同一条通信链路上，上述信息即使没有通信链路标识也依然可以推导得到。
应当理解，为了能够更准确的预测出网络设备之间的通信连接关系，关于网络结构拓扑信息的内容，还可以包括其他用于描述网络设备本自身属性的信息，例如网络设备的标识、网络设备的类型，等等。
在一种可能的实现中，用户向服务器或者终端设备直接输入的信息可以不为上述第一网络信息的形式，而是输入一个用于描述目标网络的信息，而服务器或者终端设备在接收到用户输入的用于描述目标网络的拓扑结构的信息之后，可以对其进行预处理，或者其他的信息整理方式来得到上述实施例中所描述的第一网络信息。
例如可以参照图6，图6为一种第一网络信息的示意，其中，图6所示的第一网络信息可以包括通信链路1以及通信链路2，通信链路1可以包括网络设备1、网络设备2以及网络设备3，通信链路2可以包括网络设备1、网络设备4以及网络设备5。
在一种可能的实现中，所述多个网络设备包括第一网络设备和第二网络设备。
其中，第一网络设备和第二网络设备可以为目标网络中任意的两个网络设备。
其中，第一网络设备和第二网络设备可以为用户指定的。例如，用户可以指定预测目标网络中哪两个网络设备之间的通信连接关系。
在一种可能的实现中，所述第一网络信息包括所述第一网络设备和所述第二网络设备所在的通信链路的拓扑结构信息。
在一种可能的实现中，第一网络信息可以包括第一网络设备所在的通信链路的拓扑结构信息以及第二网络设备所在的通信链路的拓扑结构信息。
在一种可能的实现中，上述第一网络设备所在的通信链路的拓扑结构信息可以为用户直接输入的信息中直接得到的，或者是从用户输入的信息当中推理得到的。
在一种可能的实现中，上述第二网络设备所在的通信链路的拓扑结构信息，可以为用户直接输入的信息中直接得到的，或者是从用户输入的信息当中推理得到的。
为了能够准确的预测出第一网络设备以及第二网络设备之间的通信连接关系，可以获取到所述第一网络设备以及所述第二网络设备相关的拓扑结构信息，也就是上述所描述的第一网络设备所在的通信链路的拓扑结构信息、以及第二网络设备所在的通信链路的拓扑结构信息。
在一种可能的实现中，第一网络设备所在通信链路的拓扑结构信息可以为第一网络设备在目标网络中所有所在的通讯任务的拓扑结构信息，也可以是第一网络设备在目标网络中所在的部分通信链路的拓扑结构信息。
关于所选的通信链路的数量以及通信链路的长度，所谓长度可以理解为通信链路上所包括的网络设备的数量，通信链路的数量以及通信链路的长度的选择可以取决于预测精度和算力之间的平衡，通信链路的数量以及规模越大，则相应的算力开销越高，最终推理得到的连接关系更准确，通信链路的数量以及规模越小，则相应的算力开销越小，最终推理得到的连接关系相对不准确。
从理论上来说，当两个网络设备同时连接于一个共同的网络设备或者是连接于很多数量的网络设备上时，可以认为这两个网络设备之间存在很大的可能性，会存在通讯连接关系，当然，网络的结构很复杂，可能会存在着比上述信息更丰富的判断规则，例如两个网络设备之间可能同时连接有多个网络设备且这两这多个网络设备之间也存在着连接关系等等，上述因素都可以都会增加网络设备之间存在通信连接关系的可能性。
本申请实施例中将待预测的网络设备所在的通讯链路的拓扑结构作为后续预测网络设备之间通讯连接关系的考虑因素。而为了能够准确地表征出，待预测的网络设备所在的通讯链路的拓扑结构关系，则需要该拓扑结构关系能够包含通信链路上网络设备与待预测的网络设备之间的连接关系以及连接位置。
在一种可能的实现中，可以通过预设的编码方式来将第一网络设备以及第二网络设备所在通信链路当中的拓扑结构信息进行编码，该编码结果可以表征出第一网络设备以及第二网络设备所在通信链路当中各个网络设备相对于第一网络设备以及第二网络设备之间的连接关系以及连接位置。
接下来介绍一个对第一网络设备以及第二网络设备所在通信链路中的拓扑结构信息的编码方式：
首先可以进行物理拓扑的还原，物理拓扑的还原主要是根据已有的链路信息将设备链接关系表示为图结构，其中物理拓扑步骤的还原旨在根据已有的链路信息将设备链接关系表示为图结构，对于一些物理拓扑数据已经是以图数据方式存储，则此步骤可以省略。进一步结构特征编码指的对物理拓扑进行结构信息的编码。
在一种可能的实现中，物理拓扑还原是根据已有的链路信息将设备链接关系表示为图结构。对于每一条路径，检查该路径中的节点是否已经在图中，然后，根据Path Hop来连接在物理拓扑中应该相连的边。
以第一网络设备和第二网络设备所在的通信链路为目标网络的一个子图为例，在一种可能的实现中，结构特征编码是指选中的节点对(链路)抽取封闭子图并基于子图进行结构信息的抽取。具体是提取出以这条边的两个节点为中心的h阶邻居节点(h大于等于1)，以及这些节点之间所形成的边，从而会形成一张以这条边的两个节点为中心的一个子图，称之为封闭子图。对于每一个封闭子图，构建其图中节点的特征。通过节点标注的方法将封闭子图中的每一个节点映射到整数集中，即：
fl:V→N；
示例性的，使用的节点标注方法由以下的标准得出：
1、每一个封闭子图中的目标节点对x和y的标号都为1；
2、节点i和节点j有相同标号当且仅当d(i,x)＝d(j,x),d(i,y)＝d(j,y)；
基于上述准则，可以采用DRNL(Double-Radius Node Labeling)方法，具体的标号结果可以示例性的如图6所示：
DRNL的优点是具有确定的计算公式：
fl(i)＝1+min(dx,dy)+(d/2)[(d)2)+(d％2)-1]；
其中dx和dy分别表示节点i到目标节点x和y的距离，d＝dx+dy，/和％分别表示除法取整和除法取余。使用DRNL进行子图结构特征的编码，编码结果满足如下的性质：
1、如果d(i,x)+d(i,y)≠d(j,x)+d(j,y)，则
2、如果d(i,x)+d(i,y)＝d(j,x)+d(j,y)，则
对节点标注完成之后，子图中的每一个节点都有相应的标注，可以以独热编码的方式子图节点构建特征，得到子图的特征矩阵X1。例如可以参照图7，图7为一种节点的标注示意，参照图8，图8为一个对第一网络信息进行编码的流程示意。
从理论上来说，当两个网络设备之间的运行状态信息相似度较高时，则大概率认为两个网络设备之间存在通信连接关系的可能性越大，例如，当两个网络设备都有一个共同的网络设备，其三者之间的运行状态信息相似度都很大，则该两个网络设备之间的通信连接关系的概率较大，当然网络的结构很复杂，存在着比上述信息更丰富的判断规则，因此本申请可以将包括第一网络设备以及第二网络设备在内的多个网络设备的运行状态信息也作 为判断第一网络设备以及第二网络设备是否存在通讯连接关系所考虑的因素。
在一种可能的实现中，可以获取到目标网络的第二网络信息，其中，所述第二网络信息包括所述第一网络设备和所述第二网络设备在内的多个网络设备的运行状态信息。
在一种可能的实现中，所述第一网络设备和所述第二网络设备在内的多个网络设备，包括：除所述第一网络设备和所述第二网络设备之外的至少一个网络设备。例如，所述第一网络设备和所述第二网络设备在内的多个网络设备，可以包括目标网络所包括的全部网络设备。
在一种可能的实现中，所述运行状态信息包括如下信息的至少一种：网络设备的告警信息、网络设备的关键性能指标(key performance indicator，KPI)。
在一种可能的实现中，所述告警信息包括如下信息的至少一种：网络设备发生告警的时间、网络设备发生的告警类型。
在一种可能的实现中，告警信息可以为网络设备的告警记录，目标网络包括多个设备，每条告警记录可以包含告警的名称、告警出现的时间和告警出现的网络设备的标识。
告警的名称又可以称为告警的类型；告警出现的设备的标识的形式可以有多种，本申请实施例对此不做具体限定，例如，告警出现的设备的标识可以是编号。
应理解，告警记录通常存在于告警日志中。所以，可以基于目标网络的告警日志获取告警记录。
作为一种实现方式，可以获取目标网络在目标时间段内产生的告警记录，其中目标时间段可以根据实际需要进行设定。例如，目标时间段可以为一个月，除此之外，目标时间段还可以为20天、25天、35天、40天等。
在一种可能的实现中，网络设备的关键性能指标可以为网络设备的通信流量。
在一种可能的实现中，可以将运行状态信息映射到嵌入向量的空间当中，也就是更高维的特征空间中，进而可以通过嵌入向量的形式来表达出运行状态信息，其中映射的目的可以为：具备更高相似度运行状态信息的网络设备之间对应的嵌入向量的相似度是更高的。
接下来给出一个计算运行状态信息所对应的嵌入向量的示例性的实现方式：
在一种可能的实现中，可以基于运行状态信息之间的相似度来构建语义拓扑，类似于上述基于通信连接关系之间的网络拓扑，语义拓扑也包括多个网络设备之间的连接关系，而和上述连接关系所对应于通信连接不同的是，语义拓扑中的连接关系为具备较高相似度的运行状态信息的网络设备之间所具有的关系。
接下来以运行状态信息为告警信息为例介绍和为相似度较高的运行状态信息：
在一种可能的实现中，可以通过网络设备之间的告警共出现性，判断是否存在相应的连接关系(或者称之为语义拓扑中的边)。参照图9，图9为一个语义拓扑的构建示意。
在一种可能的实现中，告警共出现性，可以理解为网络设备之间发生告警的时间相似度，具体的语义拓扑构造过程如下：
将一个设备上所有的告警时间进行合并得到一个告警时间列表：
[t1,t2,t3,…]；
设立时间阈值t_interval，对于每一对节点对，统计两个设备的告警时间差值小于时间间隔的数目，称之为告警共出现次数；
设立频数阈值count_interval，若一个节点对的告警共出现次数小于频数阈值，则忽略这个节点对；
对于剩下的节点，视其为语义拓扑中存在的边，从而完成语义拓扑的构建过程。
在一种可能的实现中，在运行状态信息包括通信流量的情况下，由于通信流量是一个非离散量，也就是是连续量，因此可以通过通信流量随时间的变化特征来确定网络设备的运行状态之间的相似度。
在一种可能的实现中，可以根据所述第二网络信息，得到所述第一网络设备和所述第二网络设备在内的多个网络设备的多个嵌入向量；其中，运行状态信息的相似度越大的网络设备所对应的嵌入向量之间的距离越近；所述多个嵌入向量包括所述第一网络设备的第一嵌入向量、以及所述第二网络设备的第二嵌入向量。
在得到运行状态信息的语义拓扑之后，可以基于语义拓扑进行语义结构信息的编码，示例性的可以采用无监督学习的方法来得到节点的结构特征表示。通过使用常用的无监督模型Node2Vec，得到每个网络设备的嵌入向量，从而学习到语义拓扑的结构信息X2。参照图10，图10为一个基于随机游走来构建嵌入向量的示意。
502、根据所述第一网络信息以及所述第二网络信息，通过特征提取，得到所述第一网络设备以及所述第二网络设备的目标特征向量。
在一种可能的实现中，第一网络信息可以携带和第一网络设备以及第二网络设备相关的拓扑结构信息，第二网络信息可以携带第一网络设备以及第二网络设备相关的运行状态信息，拓扑结构信息和运行状态信息都与第一网络信息和第二网络信息之间是否存在通信连接关系存在关联。
本申请实施例中，可以根据所述第一网络信息以及所述第二网络信息，通过特征提取，得到所述第一网络设备以及所述第二网络设备的目标特征向量，该目标特征向量可以携带上述第一网络设备以及第二网络设备相关的拓扑结构信息以及第一网络设备以及第二网络设备相关的运行状态信息，进而可以根据所述目标特征向量，通过目标神经网络，来预测所述第一网络设备和所述第二网络设备之间的通信连接关系。
接下来介绍如何根据所述第一网络信息以及所述第二网络信息，通过特征提取，得到所述第一网络设备以及所述第二网络设备的目标特征向量：
在一种可能的实现中，所述特征提取为基于特征提取网络实现的。
在一种可能的实现中，所述特征提取网络为图神经网络GNN。
在一种可能的实现中，可以根据所述第一网络信息、所述第一嵌入向量以及所述第二嵌入向量，通过特征提取，得到所述第一网络设备以及所述第二网络设备的目标特征向量。
在一种可能的实现中，可以根据所述第一网络信息、所述语义拓扑的信息、所述第一嵌入向量以及所述第二嵌入向量，通过特征提取，得到所述第一网络设备以及所述第二网络设备的目标特征向量。
在一种可能的实现中，可以将编码后的第一网络信息，以及编码后的第二网络信息(也就是所述第一嵌入向量以及所述第二嵌入向量)进行融合，融合的方式不限于矩阵拼接。将编码后的第一网络信息，以及编码后的第二网络信息进行融合，相当于实现了多源结构 学习的特征融合，通过特征融合，使得融合后的特征中既包含物理拓扑的结构信息，也包含与语义拓扑中的结构信息。参照图11，图11为一个特征融合的示意。
在一种可能的实现中，可以将对编码后的第一网络信息，以及编码后的第二网络信息融合后的结果，以及语义拓扑的信息(作为输入的邻接矩阵，其包括语义拓扑中各个网络设备之间的连接关系)，输入到特征提取网络中。
以第一网络设备和第二网络设备所在的通信链路为目标网络的一个子图为例，对于每一张子图，用Gsub＝(V,E,X)来表示，其中V表示顶点集，E表示边集，X表示融合后的特征矩阵，它的第i行代表顶点集中第i个节点的特征，用A来表示该子图的邻接矩阵。将其作为图神经网络的输入。图神经网络用公式化表示为：
Z＝GCN(X,A)；
其中Z∈RN×d为图神经网络的输出，表示每一个子图的特征嵌入。
503、根据所述目标特征向量，通过目标神经网络，预测所述第一网络设备和所述第二网络设备之间的通信连接关系。
在一种可能的实现中，在得到目标特征向量之后，可以根据所述目标特征向量，通过目标神经网络，预测所述第一网络设备和所述第二网络设备之间的通信连接关系。
在一种可能的实现中，所述目标神经网络为全连接网络，其中，全连接网络可以为预训练好的网络，其具备根据特征向量来预测网络设备之间的通信连接关系的能力。
在一种可能的实现中，目标神经网络的输出可以为所述第一网络设备和所述第二网络设备之间存在通信连接关系的概率。
在一种可能的实现中，可以得到目标网络中任意两组网络设备对应的特征向量，进而可以通过目标神经网络来预测网络设备之间的通信连接关系。
示例性的，以目标神经网络包括MLP层为例，经过MLP层，可以得到最终的输出Y，Y的第i个元素yi表示第i张图作为输入时得到的输出。
在所述预测所述第一网络设备和所述第二网络设备之间的通信连接关系之后，将所述第一网络设备和所述第二网络设备之间的通信连接关系传递至所述用户设备。
此外，在目标网络新增了网络设备时(例如第三网络设备)，可以接收到来自用户设备的第三网络设备的第三网络信息和第四网络信息，其中所述第三网络设备不属于所述目标网络，所述第三网络信息包括所述第三网络信息与所述多个网络设备中至少一个网络设备的通信连接关系，所述第四网络设备包括所述第三网络设备的运行状态信息；和上述类似，可以根据所述第一网络信息、所述第二网络信息、所述第三网络信息和所述第四网络设备，预测所述第三网络设备和所述多个网络设备中至少一个网络设备之间的通信连接关系；并将所述第三网络设备和所述多个网络设备中至少一个网络设备之间的通信连接关系传递至所述用户设备。
从理论上来说，当两个网络设备之间的运行状态信息相似度较高时，则大概率认为两个网络设备之间存在通信连接关系的可能性越大，例如，当两个网络设备都有一个共同的网络设备，其三者之间的运行状态信息相似度都很大，则该两个网络设备之间的通信连接关系的概率较大，当然网络的结构很复杂，存在着比上述信息更丰富的判断规则，因此本 申请可以将包括第一网络设备以及第二网络设备在内的多个网络设备的运行状态信息也作为判断第一网络设备以及第二网络设备是否存在通讯连接关系所考虑的因素。
从理论上来说，当两个网络设备之间的运行状态信息相似度较高时，则大概率认为两个网络设备之间存在通信连接关系的可能性越大，例如，当两个网络设备都有一个共同的网络设备，其三者之间的运行状态信息相似度都很大，则该两个网络设备之间的通信连接关系的概率较大，当然网络的结构很复杂，存在着比上述信息更丰富的判断规则，因此本申请可以将包括第一网络设备以及第二网络设备在内的多个网络设备的运行状态信息也作为判断第一网络设备以及第二网络设备是否存在通讯连接关系所考虑的因素。
本申请实施例提供了一种连接关系预测方法，所述方法包括：获取目标网络的第一网络信息以及第二网络信息；其中，所述目标网络包括多个网络设备，所述多个网络设备包括第一网络设备和第二网络设备，所述第一网络信息包括所述第一网络设备和所述第二网络设备所在的通信链路的拓扑结构信息，所述第二网络信息包括所述第一网络设备和所述第二网络设备在内的多个网络设备的运行状态信息；根据所述第一网络信息以及所述第二网络信息，通过特征提取，得到所述第一网络设备以及所述第二网络设备的目标特征向量；根据所述目标特征向量，通过目标神经网络，预测所述第一网络设备和所述第二网络设备之间的通信连接关系。本申请将携带和第一网络设备以及第二网络设备相关的拓扑结构信息的第一网络信息，以及携带第一网络设备以及第二网络设备相关的运行状态信息的第二网络信息作为预测第一网络设备以及第二网络设备之间是否存在通信连接关系的参考，在预测网络设备之间的连接关系时，可以识别出更丰富的数据关联关系，进而增加连接关系的预测精度。
接下来结合实验，验证本申请实施例的有益效果：
在数据集上进行实验，来衡量本发明是否有效。
表二
从表二中可以看出，本申请实施例中的预测方法的效果要优于其它的链接预测方法。
以上从模型的推理过程对本申请实施例中的连接关系预测方法进行了描述，接下来从模型训练过程对本申请实施例中的连接关系预测方法进行描述：
参照图12，图12为本申请实施例提供的一种模型训练的流程示意，如图12所示，本申请实施例提供的模型训练包括：
1201、获取目标网络的第一网络信息以及第二网络信息；其中，所述目标网络包括多 个网络设备，所述多个网络设备包括第一网络设备和第二网络设备，所述第一网络信息包括所述第一网络设备和所述第二网络设备所在的通信链路的拓扑结构信息，所述第二网络信息包括所述第一网络设备和所述第二网络设备在内的多个网络设备的运行状态信息，所述目标网络指示所述第一网络设备和所述第二网络设备之间存在第一通信连接关系。
其中目标网络可以为用户设备指定的网络，在目标网络中，第一网络设备和第二网络设备之间可以存在通信连接关系(进而，第一网络设备和第二网络设备可以作为后续模型训练的正边)，第一网络设备和第二网络设备之间可以不存在通信连接关系(进而，第一网络设备和第二网络设备可以作为后续模型训练的负边)。
关于步骤1201的具体描述可以参照步骤501，这里不再赘述。
1202、根据所述第一网络信息以及所述第二网络信息，通过特征提取，得到所述第一网络设备以及所述第二网络设备的目标特征向量。
关于步骤1202的具体描述可以参照步骤502，这里不再赘述。
1203、根据所述目标特征向量，通过目标神经网络，预测所述第一网络设备和所述第二网络设备之间的第二通信连接关系。
关于步骤1203的具体描述可以参照步骤503，这里不再赘述。
1204、根据所述第一通信连接关系以及所述第二通信连接关系，确定损失，并根据所述损失更新所述目标神经网络。
其中，可以将目标神经网络输出的所述第一网络设备和所述第二网络设备之间的通信连接关系，和目标网络所指示的第一网络设备和所述第二网络设备之间的通信连接关系来构建损失，并基于该损失训练目标神经网络(以及特征提取网络)。
可选的，在训练时需要设置损失函数，考虑到这本质上是一个二分类问题，因此可以选用交叉熵损失函数：
交叉熵损失函数能够表征样本标签和预测概率之间的差值，因此在训练过程中最小化交叉熵损失函数能够尽可能的时对样本的预测概率与真实标签保持一致。
参照图13，图13为本申请实施例提供的一种连接关系预测方法的流程示意，所述方法包括：
1301、接收来自用户设备的目标网络的第一网络信息以及第二网络信息；其中，所述目标网络包括多个网络设备，所述第一网络信息包括所述多个网络设备之间通信连接的拓扑结构信息，所述第二网络信息包括所述多个网络设备的运行状态信息；所述多个网络设备包括第一网络设备和第二网络设备；
关于步骤1301的描述可以参照上述实施例中步骤501的描述，这里不再赘述。
1302、根据所述第一网络信息以及所述第二网络信息，预测所述第一网络设备和所述第二网络设备之间的通信连接关系；
关于步骤1302的描述可以参照上述实施例中步骤502和步骤503相关的描述，这里不再赘述。
1303、将所述所述第一网络设备和所述第二网络设备之间的通信连接关系传递至所述用户设备。
关于步骤1303的描述可以参照上述实施例中步骤503相关的描述，这里不再赘述。
在一种可能的实现中，所述运行状态信息包括如下信息的至少一种：
网络设备的告警信息、网络设备的关键性能指标KPI。
在一种可能的实现中，所述告警信息包括如下信息的至少一种：
网络设备发生告警的时间、网络设备发生的告警类型。
在一种可能的实现中，所述方法还包括：
接收来自用户设备的第三网络设备的第三网络信息和第四网络信息，所述第三网络设备不属于所述目标网络，所述第三网络信息包括所述第三网络信息与所述多个网络设备中至少一个网络设备的通信连接关系，所述第四网络设备包括所述第三网络设备的运行状态信息；
根据所述第一网络信息、所述第二网络信息、所述第三网络信息和所述第四网络设备，预测所述第三网络设备和所述多个网络设备中至少一个网络设备之间的通信连接关系；
将所述第三网络设备和所述多个网络设备中至少一个网络设备之间的通信连接关系传递至所述用户设备。
在图1至图13所对应的实施例的基础上，为了更好的实施本申请实施例的上述方案，下面还提供用于实施上述方案的相关设备。具体参阅图14，图14为本申请实施例提供的一种连接关系预测装置的结构示意，如图14所示，本申请实施例提供的一种连接关系预测装置1400，包括：
获取模块1401，用于获取目标网络的第一网络信息以及第二网络信息；其中，所述目标网络包括多个网络设备，所述多个网络设备包括第一网络设备和第二网络设备，所述第一网络信息包括所述第一网络设备和所述第二网络设备所在的通信链路的拓扑结构信息，所述第二网络信息包括所述第一网络设备和所述第二网络设备在内的多个网络设备的运行状态信息；
关于获取模块1401的描述，可以参照上述实施例中步骤501的描述，这里不再赘述。
特征提取模块1402，用于根据所述第一网络信息以及所述第二网络信息，通过特征提取，得到所述第一网络设备以及所述第二网络设备的目标特征向量；
关于特征提取模块1402的描述，可以参照上述实施例中步骤502的描述，这里不再赘述。
连接关系预测模块1403，用于根据所述目标特征向量，通过目标神经网络，预测所述第一网络设备和所述第二网络设备之间的通信连接关系。
关于连接关系预测模块1403的描述，可以参照上述实施例中步骤503的描述，这里不再赘述。
在一种可能的实现中，所述运行状态信息包括如下信息的至少一种：
网络设备的告警信息、网络设备的关键性能指标KPI。
在一种可能的实现中，所述告警信息包括如下信息的至少一种：
网络设备发生告警的时间、网络设备发生的告警类型。
在一种可能的实现中，所述特征提取模块，具体用于：
根据所述第二网络信息，得到所述第一网络设备和所述第二网络设备在内的多个网络设备的多个嵌入向量；其中，运行状态信息的相似度越大的网络设备所对应的嵌入向量之间的距离越近；所述多个嵌入向量包括所述第一网络设备的第一嵌入向量、以及所述第二网络设备的第二嵌入向量；
根据所述第一网络信息、所述第一嵌入向量以及所述第二嵌入向量，通过特征提取，得到所述第一网络设备以及所述第二网络设备的目标特征向量。
在一种可能的实现中，所述第一网络设备和所述第二网络设备在内的多个网络设备，包括：
除所述第一网络设备和所述第二网络设备之外的至少一个网络设备。
在一种可能的实现中，所述特征提取模块，具体用于：
根据所述第二网络信息，得到所述第一网络设备和所述第二网络设备在内的多个网络设备的语义拓扑，所述语义拓扑包括网络设备之间的语义连接关系，其中，所述运行状态信息的相似度大于阈值的网络设备之间存在所述语义连接关系；
对所述语义拓扑进行随机游走，以得到所述第一网络设备和所述第二网络设备在内的多个网络设备的多个嵌入向量。
在一种可能的实现中，所述特征提取模块，具体用于：
根据所述第一网络信息、所述语义拓扑的信息、所述第一嵌入向量以及所述第二嵌入向量，通过特征提取，得到所述第一网络设备以及所述第二网络设备的目标特征向量。
在一种可能的实现中，所述获取模块，具体用于：
获取来自用户设备的所述目标网络的第一网络信息以及第二网络信息；
所述装置还包括：
发送模块，用于在所述预测所述第一网络设备和所述第二网络设备之间的通信连接关系之后，将所述第一网络设备和所述第二网络设备之间的通信连接关系传递至所述用户设备。
在一种可能的实现中，所述特征提取为基于特征提取网络实现的，所述特征提取网络为图神经网络GNN；所述目标神经网络为全连接网络。
参阅图15，图15为本申请实施例提供的一种连接关系预测装置的结构示意，如图15所示，本申请实施例提供的一种连接关系预测装置1500，包括：
获取模块1501，用于接收来自用户设备的目标网络的第一网络信息以及第二网络信息；其中，所述目标网络包括多个网络设备，所述第一网络信息包括所述多个网络设备之间通信连接的拓扑结构信息，所述第二网络信息包括所述多个网络设备的运行状态信息；所述多个网络设备包括第一网络设备和第二网络设备；
关于获取模块1501的描述，可以参照上述实施例中步骤1301的描述，这里不再赘述。
连接关系预测模块1502，用于根据所述第一网络信息以及所述第二网络信息，预测所述第一网络设备和所述第二网络设备之间的通信连接关系；
关于连接关系预测模块1502的描述，可以参照上述实施例中步骤1302的描述，这里不再赘述。
发送模块1503，用于将所述所述第一网络设备和所述第二网络设备之间的通信连接关系传递至所述用户设备。
关于发送模块1503的描述，可以参照上述实施例中步骤1303的描述，这里不再赘述。
在一种可能的实现中，所述运行状态信息包括如下信息的至少一种：
网络设备的告警信息、网络设备的关键性能指标KPI。
在一种可能的实现中，所述告警信息包括如下信息的至少一种：
网络设备发生告警的时间、网络设备发生的告警类型。
在一种可能的实现中，所述获取模块，还用于：
接收来自用户设备的第三网络设备的第三网络信息和第四网络信息，所述第三网络设备不属于所述目标网络，所述第三网络信息包括所述第三网络信息与所述多个网络设备中至少一个网络设备的通信连接关系，所述第四网络设备包括所述第三网络设备的运行状态信息；
所述连接关系预测模块，还用于：根据所述第一网络信息、所述第二网络信息、所述第三网络信息和所述第四网络设备，预测所述第三网络设备和所述多个网络设备中至少一个网络设备之间的通信连接关系；
所述发送模块，还用于：将所述第三网络设备和所述多个网络设备中至少一个网络设备之间的通信连接关系传递至所述用户设备。
接下来介绍本申请实施例提供的一种执行设备，请参阅图16，图16为本申请实施例提供的执行设备的一种结构示意图，执行设备1600具体可以表现为虚拟现实VR设备、手机、平板、笔记本电脑、智能穿戴设备、监控数据处理设备或服务器等，此处不做限定。具体的，执行设备1600包括：接收器1601、发射器1602、处理器1603和存储器1604(其中执行设备1600中的处理器1603的数量可以一个或多个，图16中以一个处理器为例)，其中，处理器1603可以包括应用处理器16031和通信处理器16032。在本申请的一些实施例中，接收器1601、发射器1602、处理器1603和存储器1604可通过总线或其它方式连接。
存储器1604可以包括只读存储器和随机存取存储器，并向处理器1603提供指令和数据。存储器1604的一部分还可以包括非易失性随机存取存储器(non-volatile random access memory，NVRAM)。存储器1604存储有处理器和操作指令、可执行模块或者数据结构，或者它们的子集，或者它们的扩展集，其中，操作指令可包括各种操作指令，用于实现各种操作。
处理器1603控制执行设备的操作。具体的应用中，执行设备的各个组件通过总线系统耦合在一起，其中总线系统除包括数据总线之外，还可以包括电源总线、控制总线和状态信号总线等。但是为了清楚说明起见，在图中将各种总线都称为总线系统。
上述本申请实施例揭示的方法可以应用于处理器1603中，或者由处理器1603实现。处理器1603可以是一种集成电路芯片，具有信号的处理能力。在实现过程中，上述方法的各步骤可以通过处理器1603中的硬件的集成逻辑电路或者软件形式的指令完成。上述的处理器1603可以是通用处理器、数字信号处理器(digital signal processing，DSP)、微处理器或微控制器，还可进一步包括专用集成电路(application specific integrated circuit，ASIC)、现场可编程门阵列(field-programmable gate array，FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件。该处理器1603可以实现或者执行本申请实施例中的公开的各方法、步骤及逻辑框图。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。结合本申请实施例所公开的方法的步骤可以直接体现为硬件译码处理器执行完成，或者用译码处理器中的硬件及软件模块组合执行完成。软件模块可以位于随机存储器，闪存、只读存储器，可编程只读存储器或者电可擦写可编程存储器、寄存器等本领域成熟的存储介质中。该存储介质位于存储器1604，处理器1603读取存储器1604中的信息，结合其硬件完成上述方法的步骤。
接收器1601可用于接收输入的数字或字符信息，以及产生与执行设备的相关设置以及功能控制有关的信号输入。发射器1602可用于通过第一接口输出数字或字符信息；发射器1602还可用于通过第一接口向磁盘组发送指令，以修改磁盘组中的数据；发射器1602还可以包括显示屏等显示设备。
本申请实施例中，在一种情况下，处理器1603，用于执行上述图5、图13的连接关系预测方法。
本申请实施例还提供了一种训练设备，请参阅图17，图17是本申请实施例提供的训练设备一种结构示意图，训练设备1700上可以部署有图17对应实施例中所描述的图像处理设备，用于实现图18对应实施例中数据处理设备的功能，具体的，训练设备1700由一个或多个服务器实现，训练设备1700可因配置或性能不同而产生比较大的差异，可以包括一个或一个以上中央处理器(central processing units，CPU)1717(例如，一个或一个以上处理器)和存储器1732，一个或一个以上存储应用程序1742或数据1744的存储介质1730(例如一个或一个以上海量存储设备)。其中，存储器1732和存储介质1730可以是短暂存储或持久存储。存储在存储介质1730的程序可以包括一个或一个以上模块(图示没标出)，每个模块可以包括对训练设备中的一系列指令操作。更进一步地，中央处理器1717可以设置为与存储介质1730通信，在训练设备1700上执行存储介质1730中的一系列指令操作。
训练设备1700还可以包括一个或一个以上电源1726，一个或一个以上有线或无线网络接口1750，一个或一个以上输入输出接口1758；或，一个或一个以上操作系统1741，例如Windows ServerTM，Mac OS XTM，UnixTM,LinuxTM，FreeBSDTM等等。
本申请实施例中，中央处理器1717，用于执行通过上述图12的模型训练方法。
本申请实施例中还提供一种包括计算机程序产品，当其在计算机上运行时，使得计算机执行如前述执行设备所执行的步骤，或者，使得计算机执行如前述训练设备所执行的步骤。
本申请实施例中还提供一种计算机可读存储介质，该计算机可读存储介质中存储有用于进行信号处理的程序，当其在计算机上运行时，使得计算机执行如前述执行设备所执行的步骤，或者，使得计算机执行如前述训练设备所执行的步骤。
本申请实施例提供的执行设备、训练设备或终端设备具体可以为芯片，芯片包括：处理单元和通信单元，所述处理单元例如可以是处理器，所述通信单元例如可以是输入/输出接口、管脚或电路等。该处理单元可执行存储单元存储的计算机执行指令，以使执行设备内的芯片执行上述实施例描述的数据处理方法，或者，以使训练设备内的芯片执行上述实施例描述的数据处理方法。可选地，所述存储单元为所述芯片内的存储单元，如寄存器、缓存等，所述存储单元还可以是所述无线接入设备端内的位于所述芯片外部的存储单元，如只读存储器(read-only memory，ROM)或可存储静态信息和指令的其他类型的静态存储设备，随机存取存储器(random access memory，RAM)等。
具体的，请参阅图18，图18为本申请实施例提供的芯片的一种结构示意图，所述芯片可以表现为神经网络处理器NPU 1800，NPU 1800作为协处理器挂载到主CPU(Host CPU)上，由Host CPU分配任务。NPU的核心部分为运算电路1803，通过控制器1804控制运算电路1803提取存储器中的矩阵数据并进行乘法运算。
在一些实现中，运算电路1803内部包括多个处理单元(Process Engine,PE)。在一些实现中，运算电路1803是二维脉动阵列。运算电路1803还可以是一维脉动阵列或者能够执行例如乘法和加法这样的数学运算的其它电子线路。在一些实现中，运算电路1803是通用的矩阵处理器。
举例来说，假设有输入矩阵A，权重矩阵B，输出矩阵C。运算电路从权重存储器1802中取矩阵B相应的数据，并缓存在运算电路中每一个PE上。运算电路从输入存储器1801中取矩阵A数据与矩阵B进行矩阵运算，得到的矩阵的部分结果或最终结果，保存在累加器(accumulator)1808中。
统一存储器1806用于存放输入数据以及输出数据。权重数据直接通过存储单元访问控制器(Direct Memory Access Controller，DMAC)1805，DMAC被搬运到权重存储器1802中。输入数据也通过DMAC被搬运到统一存储器1806中。
BIU为Bus Interface Unit即，总线接口单元1810，用于AXI总线与DMAC和取指存储器(Instruction Fetch Buffer，IFB)1809的交互。
总线接口单元1810(Bus Interface Unit，简称BIU)，用于取指存储器1809从外部存储器获取指令，还用于存储单元访问控制器1805从外部存储器获取输入矩阵A或者权重矩阵B的原数据。
DMAC主要用于将外部存储器DDR中的输入数据搬运到统一存储器1806或将权重数据搬运到权重存储器1802中或将输入数据数据搬运到输入存储器1801中。
向量计算单元1807包括多个运算处理单元，在需要的情况下，对运算电路的输出做进一步处理，如向量乘，向量加，指数运算，对数运算，大小比较等等。主要用于神经网络中非卷积/全连接层网络计算，如Batch Normalization(批归一化)，像素级求和，对特征平 面进行上采样等。
在一些实现中，向量计算单元1807能将经处理的输出的向量存储到统一存储器1806。例如，向量计算单元1807可以将线性函数；或，非线性函数应用到运算电路1803的输出，例如对卷积层提取的特征平面进行线性插值，再例如累加值的向量，用以生成激活值。在一些实现中，向量计算单元1807生成归一化的值、像素级求和的值，或二者均有。在一些实现中，处理过的输出的向量能够用作到运算电路1803的激活输入，例如用于在神经网络中的后续层中的使用。
控制器1804连接的取指存储器(instruction fetch buffer)1809，用于存储控制器1804使用的指令；
统一存储器1806，输入存储器1801，权重存储器1802以及取指存储器1809均为On-Chip存储器。外部存储器私有于该NPU硬件架构。
其中，上述任一处提到的处理器，可以是一个通用中央处理器，微处理器，ASIC，或一个或多个用于控制上述程序执行的集成电路。
另外需说明的是，以上所描述的装置实施例仅仅是示意性的，其中所述作为分离部件说明的单元可以是或者也可以不是物理上分开的，作为单元显示的部件可以是或者也可以不是物理单元，即可以位于一个地方，或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部模块来实现本实施例方案的目的。另外，本申请提供的装置实施例附图中，模块之间的连接关系表示它们之间具有通信连接，具体可以实现为一条或多条通信总线或信号线。
通过以上的实施方式的描述，所属领域的技术人员可以清楚地了解到本申请可借助软件加必需的通用硬件的方式来实现，当然也可以通过专用硬件包括专用集成电路、专用CPU、专用存储器、专用元器件等来实现。一般情况下，凡由计算机程序完成的功能都可以很容易地用相应的硬件来实现，而且，用来实现同一功能的具体硬件结构也可以是多种多样的，例如模拟电路、数字电路或专用电路等。但是，对本申请而言更多情况下软件程序实现是更佳的实施方式。基于这样的理解，本申请的技术方案本质上或者说对现有技术做出贡献的部分可以以软件产品的形式体现出来，该计算机软件产品存储在可读取的存储介质中，如计算机的软盘、U盘、移动硬盘、ROM、RAM、磁碟或者光盘等，包括若干指令用以使得一台计算机设备(可以是个人计算机，训练设备，或者网络设备等)执行本申请各个实施例所述的方法。
在上述实施例中，可以全部或部分地通过软件、硬件、固件或者其任意组合来实现。当使用软件实现时，可以全部或部分地以计算机程序产品的形式实现。
所述计算机程序产品包括一个或多个计算机指令。在计算机上加载和执行所述计算机程序指令时，全部或部分地产生按照本申请实施例所述的流程或功能。所述计算机可以是通用计算机、专用计算机、计算机网络、或者其他可编程装置。所述计算机指令可以存储在计算机可读存储介质中，或者从一个计算机可读存储介质向另一计算机可读存储介质传输，例如，所述计算机指令可以从一个网站站点、计算机、训练设备或数据中心通过有线(例如同轴电缆、光纤、数字用户线(DSL))或无线(例如红外、无线、微波等)方式向 另一个网站站点、计算机、训练设备或数据中心进行传输。所述计算机可读存储介质可以是计算机能够存储的任何可用介质或者是包含一个或多个可用介质集成的训练设备、数据中心等数据存储设备。所述可用介质可以是磁性介质，(例如，软盘、硬盘、磁带)、光介质(例如，DVD)、或者半导体介质(例如固态硬盘(Solid State Disk，SSD))等。

Claims (31)

  1. 一种连接关系预测方法，其特征在于，所述方法包括：
    获取目标网络的第一网络信息以及第二网络信息；其中，所述目标网络包括多个网络设备，所述多个网络设备包括第一网络设备和第二网络设备，所述第一网络信息包括所述第一网络设备和所述第二网络设备所在的通信链路的拓扑结构信息，所述第二网络信息包括所述第一网络设备和所述第二网络设备在内的多个网络设备的运行状态信息；
    根据所述第一网络信息以及所述第二网络信息，通过特征提取，得到所述第一网络设备以及所述第二网络设备的目标特征向量；
    根据所述目标特征向量，通过目标神经网络，预测所述第一网络设备和所述第二网络设备之间的通信连接关系。
  2. 根据权利要求1所述的方法，其特征在于，所述运行状态信息包括如下信息的至少一种：
    网络设备的告警信息、网络设备的关键性能指标KPI。
  3. 根据权利要求1或2所述的方法，其特征在于，所述告警信息包括如下信息的至少一种：
    网络设备发生告警的时间、网络设备发生的告警类型。
  4. 根据权利要求1至3任一所述的方法，其特征在于，所述根据所述第一网络信息以及所述第二网络信息，通过特征提取，得到所述第一网络设备以及所述第二网络设备的目标特征向量，包括：
    根据所述第二网络信息，得到所述第一网络设备和所述第二网络设备在内的多个网络设备的多个嵌入向量；所述多个嵌入向量包括所述第一网络设备的第一嵌入向量、以及所述第二网络设备的第二嵌入向量；
    根据所述第一网络信息、所述第一嵌入向量以及所述第二嵌入向量，通过特征提取，得到所述第一网络设备以及所述第二网络设备的目标特征向量。
  5. 根据权利要求4所述的方法，其特征在于，运行状态信息的相似度越大的网络设备所对应的嵌入向量之间的距离越近。
  6. 根据权利要求1至5任一所述的方法，其特征在于，所述第一网络设备和所述第二网络设备在内的多个网络设备，包括：
    除所述第一网络设备和所述第二网络设备之外的至少一个网络设备。
  7. 根据权利要求4至6任一所述的方法，其特征在于，所述根据所述第二网络信息，得到所述第一网络设备和所述第二网络设备在内的多个网络设备的多个嵌入向量，包括：
    根据所述第二网络信息，得到所述第一网络设备和所述第二网络设备在内的多个网络设备的语义拓扑，所述语义拓扑包括网络设备之间的语义连接关系，其中，所述运行状态信息的相似度大于阈值的网络设备之间存在所述语义连接关系；
    对所述语义拓扑进行随机游走，以得到所述第一网络设备和所述第二网络设备在内的多个网络设备的多个嵌入向量。
  8. 根据权利要求4至7任一所述的方法，其特征在于，所述根据所述第一网络信息、所述第一嵌入向量以及所述第二嵌入向量，通过特征提取，得到所述第一网络设备以及所述第二网络设备的目标特征向量，包括：
    根据所述第一网络信息、所述语义拓扑的信息、所述第一嵌入向量以及所述第二嵌入向量，通过特征提取，得到所述第一网络设备以及所述第二网络设备的目标特征向量。
  9. 根据权利要求1至8任一所述的方法，其特征在于，所述获取目标网络的第一网络信息以及第二网络信息，包括：
    获取来自用户设备的所述目标网络的第一网络信息以及第二网络信息；
    所述方法还包括：
    在所述预测所述第一网络设备和所述第二网络设备之间的通信连接关系之后，将所述第一网络设备和所述第二网络设备之间的通信连接关系传递至所述用户设备。
  10. 根据权利要求1至9任一所述的方法，其特征在于，所述特征提取为基于特征提取网络实现的，所述特征提取网络为图神经网络GNN；所述目标神经网络为全连接网络。
  11. 一种连接关系预测方法，其特征在于，所述方法包括：
    接收来自用户设备的目标网络的第一网络信息以及第二网络信息；其中，所述目标网络包括多个网络设备，所述第一网络信息包括所述多个网络设备之间通信连接的拓扑结构信息，所述第二网络信息包括所述多个网络设备的运行状态信息；所述多个网络设备包括第一网络设备和第二网络设备；
    根据所述第一网络信息以及所述第二网络信息，预测所述第一网络设备和所述第二网络设备之间的通信连接关系；
    将所述所述第一网络设备和所述第二网络设备之间的通信连接关系传递至所述用户设备。
  12. 根据权利要求11所述的方法，其特征在于，所述运行状态信息包括如下信息的至少一种：
    网络设备的告警信息、网络设备的关键性能指标KPI。
  13. 根据权利要求11或12所述的方法，其特征在于，所述告警信息包括如下信息的至 少一种：
    网络设备发生告警的时间、网络设备发生的告警类型。
  14. 根据权利要求11至13任一所述的方法，其特征在于，所述方法还包括：
    接收来自用户设备的第三网络设备的第三网络信息和第四网络信息，所述第三网络设备不属于所述目标网络，所述第三网络信息包括所述第三网络信息与所述多个网络设备中至少一个网络设备的通信连接关系，所述第四网络设备包括所述第三网络设备的运行状态信息；
    根据所述第一网络信息、所述第二网络信息、所述第三网络信息和所述第四网络设备，预测所述第三网络设备和所述多个网络设备中至少一个网络设备之间的通信连接关系；
    将所述第三网络设备和所述多个网络设备中至少一个网络设备之间的通信连接关系传递至所述用户设备。
  15. 一种连接关系预测装置，其特征在于，所述装置包括：
    获取模块，用于获取目标网络的第一网络信息以及第二网络信息；其中，所述目标网络包括多个网络设备，所述多个网络设备包括第一网络设备和第二网络设备，所述第一网络信息包括所述第一网络设备和所述第二网络设备所在的通信链路的拓扑结构信息，所述第二网络信息包括所述第一网络设备和所述第二网络设备在内的多个网络设备的运行状态信息；
    特征提取模块，用于根据所述第一网络信息以及所述第二网络信息，通过特征提取，得到所述第一网络设备以及所述第二网络设备的目标特征向量；
    连接关系预测模块，用于根据所述目标特征向量，通过目标神经网络，预测所述第一网络设备和所述第二网络设备之间的通信连接关系。
  16. 根据权利要求15所述的装置，其特征在于，所述运行状态信息包括如下信息的至少一种：
    网络设备的告警信息、网络设备的关键性能指标KPI。
  17. 根据权利要求15或16所述的装置，其特征在于，所述告警信息包括如下信息的至少一种：
    网络设备发生告警的时间、网络设备发生的告警类型。
  18. 根据权利要求15至17任一所述的装置，其特征在于，所述特征提取模块，具体用于：
    根据所述第二网络信息，得到所述第一网络设备和所述第二网络设备在内的多个网络设备的多个嵌入向量；所述多个嵌入向量包括所述第一网络设备的第一嵌入向量、以及所述第二网络设备的第二嵌入向量；
    根据所述第一网络信息、所述第一嵌入向量以及所述第二嵌入向量，通过特征提取， 得到所述第一网络设备以及所述第二网络设备的目标特征向量。
  19. 根据权利要求18所述的装置，其特征在于，运行状态信息的相似度越大的网络设备所对应的嵌入向量之间的距离越近。
  20. 根据权利要求15至19任一所述的装置，其特征在于，所述第一网络设备和所述第二网络设备在内的多个网络设备，包括：
    除所述第一网络设备和所述第二网络设备之外的至少一个网络设备。
  21. 根据权利要求18至20任一所述的装置，其特征在于，所述特征提取模块，具体用于：
    根据所述第二网络信息，得到所述第一网络设备和所述第二网络设备在内的多个网络设备的语义拓扑，所述语义拓扑包括网络设备之间的语义连接关系，其中，所述运行状态信息的相似度大于阈值的网络设备之间存在所述语义连接关系；
    对所述语义拓扑进行随机游走，以得到所述第一网络设备和所述第二网络设备在内的多个网络设备的多个嵌入向量。
  22. 根据权利要求18至21任一所述的装置，其特征在于，所述特征提取模块，具体用于：
    根据所述第一网络信息、所述语义拓扑的信息、所述第一嵌入向量以及所述第二嵌入向量，通过特征提取，得到所述第一网络设备以及所述第二网络设备的目标特征向量。
  23. 根据权利要求15至22任一所述的装置，其特征在于，所述获取模块，具体用于：
    获取来自用户设备的所述目标网络的第一网络信息以及第二网络信息；
    所述装置还包括：
    发送模块，用于在所述预测所述第一网络设备和所述第二网络设备之间的通信连接关系之后，将所述第一网络设备和所述第二网络设备之间的通信连接关系传递至所述用户设备。
  24. 根据权利要求15至23任一所述的装置，其特征在于，所述特征提取为基于特征提取网络实现的，所述特征提取网络为图神经网络GNN；所述目标神经网络为全连接网络。
  25. 一种连接关系预测装置，其特征在于，所述装置包括：
    获取模块，用于接收来自用户设备的目标网络的第一网络信息以及第二网络信息；其中，所述目标网络包括多个网络设备，所述第一网络信息包括所述多个网络设备之间通信连接的拓扑结构信息，所述第二网络信息包括所述多个网络设备的运行状态信息；所述多个网络设备包括第一网络设备和第二网络设备；
    连接关系预测模块，用于根据所述第一网络信息以及所述第二网络信息，预测所述第一网络设备和所述第二网络设备之间的通信连接关系；
    发送模块，用于将所述所述第一网络设备和所述第二网络设备之间的通信连接关系传递至所述用户设备。
  26. 根据权利要求25所述的装置，其特征在于，所述运行状态信息包括如下信息的至少一种：
    网络设备的告警信息、网络设备的关键性能指标KPI。
  27. 根据权利要求25或26所述的装置，其特征在于，所述告警信息包括如下信息的至少一种：
    网络设备发生告警的时间、网络设备发生的告警类型。
  28. 根据权利要求25至27任一所述的装置，其特征在于，所述获取模块，还用于：
    接收来自用户设备的第三网络设备的第三网络信息和第四网络信息，所述第三网络设备不属于所述目标网络，所述第三网络信息包括所述第三网络信息与所述多个网络设备中至少一个网络设备的通信连接关系，所述第四网络设备包括所述第三网络设备的运行状态信息；
    所述连接关系预测模块，还用于：根据所述第一网络信息、所述第二网络信息、所述第三网络信息和所述第四网络设备，预测所述第三网络设备和所述多个网络设备中至少一个网络设备之间的通信连接关系；
    所述发送模块，还用于：将所述第三网络设备和所述多个网络设备中至少一个网络设备之间的通信连接关系传递至所述用户设备。
  29. 一种连接关系预测装置，其特征在于，所述装置包括存储器和处理器；所述存储器存储有代码，所述处理器被配置为获取所述代码，并执行如权利要求1至14任一所述的方法。
  30. 一种计算机存储介质，其特征在于，所述计算机存储介质存储有一个或多个指令，所述指令在由一个或多个计算机执行时使得所述一个或多个计算机实施权利要求1至14任一所述的方法。
  31. 一种包含指令的计算机程序产品，其特征在于，当所述计算机程序产品在计算机上运行时，使得所述计算机执行所述权利要求1至14任一所述的方法。
PCT/CN2023/073707 2022-01-30 2023-01-29 一种连接关系预测方法及相关设备 WO2023143570A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202210114738.5 2022-01-30
CN202210114738.5A CN114900435B (zh) 2022-01-30 2022-01-30 一种连接关系预测方法及相关设备

Publications (1)

Publication Number Publication Date
WO2023143570A1 true WO2023143570A1 (zh) 2023-08-03

Family

ID=82716151

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2023/073707 WO2023143570A1 (zh) 2022-01-30 2023-01-29 一种连接关系预测方法及相关设备

Country Status (2)

Country Link
CN (1) CN114900435B (zh)
WO (1) WO2023143570A1 (zh)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114900435B (zh) * 2022-01-30 2023-12-08 华为技术有限公司 一种连接关系预测方法及相关设备

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108712286A (zh) * 2018-05-22 2018-10-26 华为技术有限公司 网络拓扑结构的确定方法、装置和存储介质
US20200366690A1 (en) * 2019-05-16 2020-11-19 Nec Laboratories America, Inc. Adaptive neural networks for node classification in dynamic networks
CN113032638A (zh) * 2021-03-23 2021-06-25 深圳大学 一种网络链路预测方法及装置
CN113315655A (zh) * 2021-05-24 2021-08-27 恒隆通信技术有限公司 智能组网环境的信息配置方法及智能组网系统
CN113869496A (zh) * 2021-09-30 2021-12-31 华为技术有限公司 一种神经网络的获取方法、数据处理方法以及相关设备
CN114900435A (zh) * 2022-01-30 2022-08-12 华为技术有限公司 一种连接关系预测方法及相关设备

Family Cites Families (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105183796A (zh) * 2015-08-24 2015-12-23 同济大学 一种基于聚类的分布式链路预测方法
CN109194509B (zh) * 2018-08-28 2022-03-01 浙江工业大学 一种基于路径强弱关系和社区信息的预测网络连边的方法
CN111953507B (zh) * 2019-05-17 2022-06-28 中移(苏州)软件技术有限公司 一种网络拓扑发现方法、装置以及计算机存储介质
US11146463B2 (en) * 2019-06-05 2021-10-12 Cisco Technology, Inc. Predicting network states for answering what-if scenario outcomes
US11032150B2 (en) * 2019-06-17 2021-06-08 International Business Machines Corporation Automatic prediction of behavior and topology of a network using limited information
CN112653586B (zh) * 2019-10-12 2022-04-19 苏州工业园区测绘地理信息有限公司 基于全链路监控的时空大数据平台应用性能管理方法
CN111368074B (zh) * 2020-02-24 2022-06-10 西安电子科技大学 一种基于网络结构和文本信息的链路预测方法
CN113704506A (zh) * 2021-04-06 2021-11-26 腾讯科技（深圳）有限公司 一种媒体内容排重方法和相关装置
CN113988464A (zh) * 2021-11-17 2022-01-28 国家电网有限公司客户服务中心 基于图神经网络的网络链路属性关系预测方法及设备

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108712286A (zh) * 2018-05-22 2018-10-26 华为技术有限公司 网络拓扑结构的确定方法、装置和存储介质
US20200366690A1 (en) * 2019-05-16 2020-11-19 Nec Laboratories America, Inc. Adaptive neural networks for node classification in dynamic networks
CN113032638A (zh) * 2021-03-23 2021-06-25 深圳大学 一种网络链路预测方法及装置
CN113315655A (zh) * 2021-05-24 2021-08-27 恒隆通信技术有限公司 智能组网环境的信息配置方法及智能组网系统
CN113869496A (zh) * 2021-09-30 2021-12-31 华为技术有限公司 一种神经网络的获取方法、数据处理方法以及相关设备
CN114900435A (zh) * 2022-01-30 2022-08-12 华为技术有限公司 一种连接关系预测方法及相关设备

Also Published As

Publication number Publication date
CN114900435A (zh) 2022-08-12
CN114900435B (zh) 2023-12-08

Similar Documents

Publication Publication Date Title
US20230196117A1 (en) Training method for semi-supervised learning model, image processing method, and device
CN109919316B (zh) 获取网络表示学习向量的方法、装置和设备及存储介质
CN110837602B (zh) 基于表示学习和多模态卷积神经网络的用户推荐方法
CN112990211B (zh) 一种神经网络的训练方法、图像处理方法以及装置
WO2019144892A1 (zh) 数据处理方法、装置、存储介质和电子装置
CN113011282A (zh) 图数据处理方法、装置、电子设备及计算机存储介质
WO2022171066A1 (zh) 基于物联网设备的任务分配方法、网络训练方法及装置
WO2020038100A1 (zh) 一种特征关系推荐方法及装置、一种计算设备及存储介质
CN112232524B (zh) 多标签信息的识别方法、装置、电子设备和可读存储介质
CN113159283A (zh) 一种基于联邦迁移学习的模型训练方法及计算节点
CN112200266B (zh) 基于图结构数据的网络训练方法、装置以及节点分类方法
WO2024032096A1 (zh) 反应物分子的预测方法、训练方法、装置以及电子设备
CN113761250A (zh) 模型训练方法、商户分类方法及装置
CN113162787B (zh) 电信网络中故障定位的方法、节点分类方法以及相关设备
WO2023217127A1 (zh) 一种因果关系确定方法及相关设备
CN111931002A (zh) 一种匹配方法以及相关设备
CN114639483A (zh) 一种基于图神经网络的电子病历检索方法及装置
WO2023143570A1 (zh) 一种连接关系预测方法及相关设备
US20230289572A1 (en) Neural network structure determining method and apparatus
CN113626612A (zh) 一种基于知识图谱推理的预测方法和系统
CN113191479A (zh) 联合学习的方法、系统、节点及存储介质
Chan et al. Deep neural networks in the cloud: Review, applications, challenges and research directions
WO2024067373A1 (zh) 一种数据处理方法及相关装置
WO2023197857A1 (zh) 一种模型切分方法及其相关设备
CN116992151A (zh) 一种基于双塔图卷积神经网络的在线课程推荐方法

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 23746447

Country of ref document: EP

Kind code of ref document: A1