WO2022105374A1 - 信息处理方法、模型的生成及训练方法、电子设备和介质 - Google Patents

信息处理方法、模型的生成及训练方法、电子设备和介质 Download PDF

Info

Publication number
WO2022105374A1
WO2022105374A1 PCT/CN2021/116968 CN2021116968W WO2022105374A1 WO 2022105374 A1 WO2022105374 A1 WO 2022105374A1 CN 2021116968 W CN2021116968 W CN 2021116968W WO 2022105374 A1 WO2022105374 A1 WO 2022105374A1
Authority
WO
WIPO (PCT)
Prior art keywords
transmission
network
model
information
node
Prior art date
Application number
PCT/CN2021/116968
Other languages
English (en)
French (fr)
Inventor
王大江
叶友道
杨玺坤
栗伟清
王振宇
Original Assignee
中兴通讯股份有限公司
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 中兴通讯股份有限公司 filed Critical 中兴通讯股份有限公司
Priority to EP21893525.2A priority Critical patent/EP4243360A4/en
Priority to JP2023528716A priority patent/JP7544976B2/ja
Priority to US18/253,130 priority patent/US20240015077A1/en
Publication of WO2022105374A1 publication Critical patent/WO2022105374A1/zh

Links

Images

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
    • H04L41/14—Network analysis or design
    • H04L41/147—Network analysis or design for predicting network behaviour
    • 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
    • H04L41/145—Network analysis or design involving simulating, designing, planning or modelling of a network
    • 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
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/0464—Convolutional networks [CNN, ConvNet]
    • 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
    • G06N3/048—Activation functions
    • 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
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • 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
    • G06N3/09—Supervised learning
    • 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
    • H04L41/08—Configuration management of networks or network elements
    • H04L41/0803—Configuration setting
    • 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
    • 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
    • H04L41/142—Network analysis or design using statistical or mathematical 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
    • H04L41/16—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks using machine learning or artificial intelligence
    • 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
    • H04L41/50—Network service management, e.g. ensuring proper service fulfilment according to agreements
    • H04L41/5003—Managing SLA; Interaction between SLA and QoS
    • H04L41/5009—Determining service level performance parameters or violations of service level contracts, e.g. violations of agreed response time or mean time between failures [MTBF]

Definitions

  • the embodiments of the present application relate to, but are not limited to, the field of network technologies, and in particular, relate to an information processing method, a method for generating a network model, a method for training a network model, an electronic device, and a computer-readable storage medium.
  • the network topology becomes more and more complex.
  • the current transmission performance prediction of services in the network mainly relies on the experience of operation and maintenance personnel, and the degree of intelligence is low.
  • the factors affecting the transmission performance prediction of services include time, space, social environment, etc. These factors are numerous and complex, which makes it difficult for simple machine learning algorithms to cope with the realization of the transmission performance prediction function affected by multi-dimensional factors. Therefore, the prediction of the transmission performance of the service in the network cannot be effectively realized.
  • Embodiments of the present application provide an information processing method, a method for generating a network model, a method for training a network model, an electronic device, and a computer-readable storage medium, which can effectively predict the transmission performance of service information in a network.
  • an embodiment of the present application provides an information processing method, including: acquiring a transmission parameter of service information in a network, wherein the transmission parameter represents the transmission performance of the service information in the network; using a pre-trained The network model performs transmission performance prediction on the transmission parameters, wherein the network model includes a graph convolution network (Graph Convolution Network, GCN) model, and the GCN model is obtained according to the transmission path information of service information in the network.
  • GCN Graph Convolution Network
  • an embodiment of the present application further provides a method for generating a network model, including: acquiring transmission path information of service information in a network; generating a GCN model according to the transmission path information, wherein the GCN model includes a method for An input node for inputting transmission parameters of the service information, where the input node corresponds to the transmission path information, and the transmission parameters represent the transmission performance of the service information in the network.
  • an embodiment of the present application further provides a method for training a network model, including: acquiring transmission parameters of service information in the network, where the transmission parameters represent the transmission performance of the service information in the network; The network model is trained using the transmission parameters, wherein the network model includes a GCN model, and the GCN model is obtained according to the transmission path information of the service information in the network.
  • an embodiment of the present application further provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and running on the processor, the processor implements the above when executing the computer program The information processing method of the first aspect.
  • an embodiment of the present application further provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and running on the processor, the processor implements the above when executing the computer program The method for generating a network model of the second aspect.
  • an embodiment of the present application further provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and running on the processor, the processor implements the above when executing the computer program The training method of the network model of the third aspect.
  • the embodiments of the present application further provide a computer-readable storage medium storing computer-executable instructions, where the computer-executable instructions are used to execute the information processing method of the first aspect as described above, or to execute the above-mentioned information processing method.
  • FIG. 1 is a schematic diagram of a network topology for executing an information processing method provided by an embodiment of the present application
  • FIG. 3 is a schematic structural diagram of a GCN model provided by an embodiment of the present application.
  • FIG. 4 is a flowchart of transmission performance prediction in an information processing method provided by another embodiment of the present application.
  • FIG. 5 is a schematic diagram of a prediction model system for predicting transmission performance of transmission parameters provided by an example of the present application
  • FIG. 6 is a flowchart of a method for generating a network model provided by an embodiment of the present application.
  • FIG. 7 is a flowchart of generating a GCN model in a method for generating a network model provided by another embodiment of the present application.
  • FIG. 8 is a schematic diagram of a service network virtual topology provided by a specific example of the present application.
  • FIG. 9 is a schematic diagram of node linking processing provided by a specific example of the present application.
  • FIG. 10 is a schematic diagram of a network topology provided by a specific example of the present application.
  • FIG. 11 is a schematic diagram of a GCN model constructed according to the network topology of FIG. 10 provided by another specific example of the present application;
  • FIG. 12 is a flowchart of a training method of a network model provided by an embodiment of the present application.
  • FIG. 13 is a flowchart of training a network model in a method for training a network model provided by another embodiment of the present application.
  • the present application provides an information processing method, a method for generating a network model, a method for training a network model, an electronic device, and a computer-readable storage medium, wherein the information processing method includes: acquiring transmission parameters of service information in a network, wherein the transmission The parameters represent the transmission performance of the service information in the network; the pre-trained network model is used to predict the transmission performance of the transmission parameters, wherein the network model includes the GCN model, and the GCN model is obtained according to the transmission path information of the service information in the network. Since the GCN model is obtained according to the transmission path information of the service information in the network, the GCN model can conform to the network topology for transmitting the service information. Therefore, by using the network model including the GCN model, the service information in the network It can accurately and effectively realize the prediction of the transmission performance of the service information in the network.
  • FIG. 1 is a schematic diagram of a network topology for executing an information processing method provided by an embodiment of the present application.
  • the network topology includes a network controller (not shown in the figure) and a plurality of nodes (node A, node B, . . . , node S in FIG. 1 ), these nodes
  • the network topology shown in FIG. 1 is formed by link connection, and the network controller is connected with each node.
  • the network controller can obtain information such as node information and service information reported by each node, so as to form a network topology model according to the node information, and can plan and configure the transmission path of service information according to the network topology model. After the network controller completes the planning and configuration of the transmission path of the service information and obtains the forwarding policy, the network controller will deliver the forwarding policy to the corresponding node, so that the nodes in the network can transmit the service information according to the forwarding policy. For example, in the network topology shown in FIG.
  • a first service transmission path from node A to node L there are the following service transmission paths: a first service transmission path from node A to node L, a second service transmission path from node A to node P, and a second service transmission path from node A to node A
  • the network controller can also predict the transmission performance of the service information according to the node information and service information uploaded by each node, and can make corresponding predictions on the results. Optimized processing.
  • each node can obtain node information advertised by other nodes in the network, so that network topology information can be obtained according to the node information.
  • a corresponding network topology model can be formed according to the network topology information, and the transmission path of the service information can be planned and configured according to the transmission performance requirements of the service information to be transmitted, and a system that meets the requirements can be obtained. forwarding strategy, and then forward the service information according to the forwarding strategy.
  • the head node that sends the service information can also predict the transmission performance of the service information, and can perform corresponding optimization processing on the prediction result.
  • the network topology and application scenarios described in the embodiments of the present application are for the purpose of illustrating the technical solutions of the embodiments of the present application more clearly, and do not constitute a limitation on the technical solutions provided by the embodiments of the present application.
  • the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.
  • FIG. 1 does not constitute a limitation on the embodiments of the present application, and may include more or less components than those shown in the figure, or combine some components, or different Component placement.
  • the network controller or the head node that sends the service information can call its stored information processing program to execute the information processing method.
  • FIG. 2 is a flowchart of an information processing method provided by an embodiment of the present application.
  • the information processing method is applied to a network controller or a head node that sends service information.
  • the information processing method includes but is not limited to the following steps: S110 and step S120.
  • Step S110 acquiring transmission parameters of the service information in the network, wherein the transmission parameters represent the transmission performance of the service information in the network.
  • the transmission parameters represent the transmission performance of the service information in the network, and the transmission parameters include but are not limited to service flow, optical signal-to-noise ratio (Optical Signal Noise Ratio, OSNR), bit error rate, and optical output power, etc.
  • OSNR optical Signal Noise Ratio
  • the transmission parameters of the service information can be obtained according to the node capability information of the head node, the node capability information of the tail node, and the node capability information of the intermediate nodes passed from the head node to the tail node.
  • the transmission parameter of the service information may be a transmission parameter obtained by prediction, or may be a transmission parameter obtained by real-time detection from the network, which is not specifically limited in this embodiment.
  • the transmission parameters of the business information are transmission parameters obtained through prediction processing
  • a prediction model constructed by a neural network, etc. can be used to analyze the node capability information of the head node, the node capability information of the tail node, and the data generated by the head node. Contents such as node capability information of intermediate nodes passed by the tail node are predicted and processed, thereby obtaining the transmission parameters of the service information.
  • Step S120 using a pre-trained network model to predict the transmission performance of the transmission parameters, wherein the network model includes a GCN model, and the GCN model is obtained according to the transmission path information of the service information in the network.
  • the GCN model since the GCN model is obtained according to the transmission path information of the service information in the network, the GCN model can be consistent with the network topology for transmitting the service information.
  • the network model processes the transmission parameters of the service information, which can more accurately and effectively realize the prediction of the transmission performance of the service information in the network.
  • the network model can be pre-trained and stored in the network controller or in the first node that sends the service information. Therefore, after obtaining the transmission parameters of the service information in the network, the network model can be used to The service information is used for transmission performance prediction. Since the network model predicts the transmission performance of the transmission parameters of business information, the training data used to train the network model is also the transmission parameters of business information. It can be understood that the training data used to train the network model can include There are various acquisition methods, which are not specifically limited in this embodiment. For example, it can be obtained by acquiring the transmission parameters of the actually transmitted service information from the network, or it can be obtained by using artificial simulation. When artificial simulation is used to obtain the training data, a virtual network topology can be constructed according to the actual network topology, and then the transmission of service information is simulated in the virtual network topology to obtain the training data.
  • the pre-trained network model including the GCN model is used to predict the transmission performance of the transmission parameters of the service information, because the GCN model is based on the transmission path of the service information in the network. Therefore, by using the network model including the GCN model to process the transmission parameters of the service information in the network, it can be more accurately and effectively realized Prediction of the transmission performance of service information in the network.
  • the GCN model since the GCN model is obtained according to the transmission path information of the service information in the network, as shown in FIG. 3 , the GCN model may include an input node for inputting transmission parameters, the input node and the transmission path information Correspondingly, the input nodes corresponding to the transmission path information with the homologous relationship or the homoclinic relationship are connected.
  • the GCN model includes input node AL, input node AP, input node AQ, input node BP, input node BL, input node BH, input node LQ, input node PQ and input node HQ, input node AL
  • the input node AP corresponds to the second service transmission path in the network topology shown in FIG. 1
  • the input node AQ corresponds to the network topology shown in FIG. 1 .
  • the third service transmission path, the input node BP corresponds to the fourth service transmission path in the network topology shown in Figure 1
  • the input node BL corresponds to the fifth service transmission path in the network topology shown in Figure 1
  • the input node BH corresponds to the sixth service transmission path in the network topology shown in Figure 1
  • the input node LQ corresponds to the seventh service transmission path in the network topology shown in Figure 1
  • the input node PQ corresponds to the network topology shown in Figure 1.
  • the eighth service transmission path of the input node HQ corresponds to the ninth service transmission path in the network topology shown in FIG. 1 .
  • the first service transmission path, the second service transmission path and the third service transmission path have the same source node (ie node A), that is, there is a homologous relationship, so the input node AL, the input node AP and the input node AQ are paired Connection, similarly, input node BP, input node BL and input node BH are connected in pairs, input node AQ, input node LQ, input node PQ and input node HQ are connected in pairs, input node AL, input node BL and input node LQ are connected Pairs are connected, the input node AP, the input node BP and the input node PQ are connected in pairs, and the input node BH is connected with the input node HQ.
  • the transmission parameter represents the transmission performance of the service information in the network
  • the transmission parameter represents the transmission performance of the service information when the service information is transmitted in the corresponding transmission path in the network
  • the GCN model obtained from the path information processes the transmission parameters of the service information, and can convert the transmission performance attributes (that is, the transmission parameters) based on the service information into the parameters input to each input node to meet the needs of the GCN model processing, so as to include
  • the network model of the GCN model can predict the transmission performance of the service information in the network.
  • the network model may also include a deep neural network (Deep Neural Network, DNN) model, and the input of the DNN model is connected to the output of the GCN model.
  • DNN Deep Neural Network
  • step S120 Using the pre-trained network model to predict the transmission performance of the transmission parameters in may include but not be limited to the following steps:
  • Step S121 using the GCN model to process the transmission parameters to obtain transmission characteristic parameters
  • Step S122 using the DNN model to process the transmission characteristic parameters to obtain a transmission performance prediction result.
  • the transmission parameters of the service information at different times can be obtained, and then, the GCN model is used to perform feature extraction on these transmission parameters according to the time dimension and the space dimension to obtain the transmission characteristic parameters, and then, these transmission characteristic parameters are used as The input of the DNN model is transmitted to the DNN model, so that the DNN model can process these transmission characteristic parameters to obtain the transmission performance prediction result for the future time of the service information.
  • the prediction model system as shown in Figure 5 first obtain the transmission parameters of the service information at each time, for example, obtain the corresponding values of n input nodes at time t and time t+1.
  • the transmission parameters of the business information that is, the input parameters of the n input nodes in the input part as shown in Figure 5, and then these input parameters are input into the GCN model processing part shown in Figure 5, so that the GCN model can These input parameters are processed to obtain the transmission characteristic parameters corresponding to the next moment.
  • the transmission characteristic parameters are input into the processing part of the DNN model as shown in Figure 5, so that the DNN model can process the transmission characteristic parameters,
  • the transmission performance prediction result of the service information at the next moment is obtained. Therefore, the transmission performance prediction result can be used as the guidance information for the optimization process to perform transmission performance optimization processing on the service information to improve the transmission performance of the service information.
  • using the GCN model to perform feature extraction on the transmission parameters according to the time dimension refers to the process of using the GCN model to perform feature extraction on the transmission parameters at different times.
  • Using the GCN model to perform feature extraction on the transmission parameters according to the spatial dimension refers to using the GCN model to perform convolution processing on the transmission parameters to obtain feature parameters with spatial characteristics.
  • y is the transmission characteristic parameter, that is, y is the prediction processing of the transmission parameter by the GCN model
  • the output parameter after; x is the transmission parameter, that is, x is the input parameter input to the GCN model;
  • ⁇ is the convolution kernel function; *g is the graph convolution operator using the spectral method, representing the product of x and ⁇ ; ⁇ is the convolution kernel parameter; L is the normalized graph Laplacian;
  • ⁇ is the activation function, such as the ReLU activation function.
  • y ⁇ R n represents the transmission characteristic parameters with temporal and spatial characteristics obtained after the GCN model performs graph convolution on the transmission parameters
  • n is the number of service information transmission in the network (that is, the transmission path for transmitting service information).
  • x ⁇ R M ⁇ n ⁇ k representing n transmission parameters input to the GCN model at M times, and each transmission parameter is a k-dimensional transmission performance attribute
  • ⁇ jt ⁇ R n representing The convolution kernel parameter of the j-th dimension transmission performance attribute of n transmission parameters at the t-th time
  • x jt ⁇ R n representing the j-th dimension transmission of the n transmission parameters input to the GCN model at the t-th time performance properties.
  • the information processing method may further include, but is not limited to, the following steps:
  • the GCN model is obtained according to the transmission path information of the service information in the network, the GCN model can be consistent with the network topology of the transmission service information.
  • the transmission performance prediction result can be used as the guidance information for the optimization process to perform transmission performance optimization processing on the service information to improve the transmission performance of the service information.
  • the transmission performance optimization processing for the service information using the transmission performance prediction result may have different processing methods, which is not specifically limited in this embodiment.
  • the service information may be adjusted in the corresponding transmission path according to the transmission performance prediction result.
  • the transmission path of the service information in the network can also be changed according to the transmission performance prediction result to achieve the purpose of optimizing the transmission performance.
  • FIG. 6 is a flowchart of a method for generating a network model provided by another embodiment of the present application.
  • the method for generating a network model is applied to a network controller or a head node that sends service information.
  • the network model The generation method includes but is not limited to the following steps:
  • Step S210 acquiring the transmission path information of the service information in the network
  • Step S220 Generate a GCN model according to the transmission path information, wherein the GCN model includes an input node for inputting transmission parameters of the service information, the input nodes correspond to the transmission path information, and the transmission parameters represent the transmission performance of the service information in the network.
  • the GCN model generated according to the transmission path information in this embodiment is the GCN model used to predict the transmission performance of the transmission parameters of the service information in the network in the embodiment shown in FIG. 2 .
  • the transmission path information of the service information in the network may include head node information and tail node information, wherein the head node information includes the address information of the head node that sends the service information, and the tail node information includes the tail node that receives the service information. address information.
  • the address information of the head node and the address information of the tail node can be obtained from the head node, so that the transmission path information of the service information can be obtained.
  • the transmission parameter represents the transmission performance of the service information in the network, and the transmission parameter includes but is not limited to service flow, OSNR, bit error rate, and optical output power.
  • the transmission parameters of the service information can be obtained according to the node capability information of the head node, the node capability information of the tail node, and the node capability information of the intermediate nodes passed from the head node to the tail node.
  • the transmission parameter of the service information may be a transmission parameter obtained by prediction, or may be a transmission parameter obtained by real-time detection from the network, which is not specifically limited in this embodiment.
  • the transmission parameters of the business information are transmission parameters obtained through prediction processing
  • a prediction model constructed by a neural network, etc. can be used to analyze the node capability information of the head node, the node capability information of the tail node, and the data generated by the head node. Contents such as node capability information of intermediate nodes passed by the tail node are predicted and processed, thereby obtaining the transmission parameters of the service information.
  • a GCN model including an input node for inputting transmission parameters of the service information can be generated according to the transmission path information, and the input node of the GCN model corresponds to the transmission path information,
  • the input node AL as shown in FIG. 3 corresponds to the first traffic transmission path as shown in FIG. 1 . Since the transmission parameters represent the transmission performance of the service information in the network, and the GCN model is obtained according to the transmission path information of the service information in the network, using the transmission parameters as the input parameters of the input node of the GCN model can be implemented more accurately and effectively. Prediction of the transmission performance of service information in the network.
  • step S220 generating the GCN model according to the transmission path information may include but not be limited to the following steps:
  • Step S221 establishing an input node according to the head node information and the tail node information
  • Step S222 Connect the input nodes corresponding to the transmission path information with the homologous relationship or the homoclinic relationship to obtain a GCN model.
  • the obtained GCN model can perform prediction processing on the transmission parameters of the service information in the corresponding transmission path, and can accurately and effectively realize the prediction of the transmission performance of the service information in the network.
  • the generation of the GCN model is described in detail below with a specific example.
  • the service network shown in Figure 8 is abstracted Virtual topology, wherein the service network virtual topology includes only the head node and tail node of each service path, and for each service path, its head node and tail node are directly connected to form a service virtual link.
  • node L can be abstractly transformed into 3 virtual paths.
  • the GCN model Since the GCN model is obtained according to the transmission path information of the service information in the network, the GCN model can conform to the network topology of the transmission service information. Therefore, by using the GCN model, the relevant prediction processing is performed on the transmission parameters of the service information in the network. , which can more accurately and effectively realize the prediction of the transmission performance of the service information in the network.
  • the network topology of multiple links between adjacent nodes can be converted into a network with a single link between adjacent nodes by adding virtual dummy nodes. Topological form, and then the GCN model can be constructed in the above-mentioned manner.
  • the entire service path can be divided into multiple service path segments according to the head node, tail node and electrical relay node, and each service path can be divided into multiple service path segments.
  • Each path segment is considered as an independent service path and virtualized.
  • y ⁇ R n represents the transmission characteristic parameters with temporal and spatial characteristics obtained after the GCN model performs graph convolution on the transmission parameters
  • n is the number of service information transmission in the network (that is, the transmission path for transmitting service information).
  • x ⁇ R M ⁇ n ⁇ k representing n transmission parameters input to the GCN model at M times, and each transmission parameter is a k-dimensional transmission performance attribute
  • ⁇ jt ⁇ R n representing The convolution kernel parameter of the j-th dimension transmission performance attribute of n transmission parameters at the t-th time
  • x jt ⁇ R n representing the j-th dimension transmission of the n transmission parameters input to the GCN model at the t-th time performance properties.
  • FIG. 12 is a flowchart of a training method for a network model provided by another embodiment of the present application.
  • the training method for the network model is applied to a network controller or a head node that sends service information.
  • the network model The training method includes but is not limited to the following steps:
  • Step S310 acquiring transmission parameters of the service information in the network, wherein the transmission parameters represent the transmission performance of the service information in the network;
  • Step S320 using the transmission parameters to train the network model, wherein the network model includes a GCN model, and the GCN model is obtained according to the transmission path information of the service information in the network.
  • the GCN model trained by using the transmission parameters of the service information in the network in this embodiment is the pre-trained GCN model in the embodiment shown in FIG. 2 , and is also implemented as shown in FIG. 6 .
  • the GCN model is generated according to the transmission path information.
  • the GCN model is obtained by using the generation method in the embodiment shown in FIG. 6 , and then the GCN model is trained by using the training method in this embodiment. After the training is completed, the GCN model in the embodiment shown in FIG. 2 can be obtained.
  • the transmission path information of the service information in the network may include head node information and tail node information, wherein the head node information includes the address information of the head node that sends the service information, and the tail node information includes the tail node that receives the service information. address information.
  • the address information of the head node and the address information of the tail node can be obtained from the head node, so that the transmission path information of the service information can be obtained.
  • the transmission parameter represents the transmission performance of the service information in the network, and the transmission parameter includes but is not limited to service flow, OSNR, bit error rate, and optical output power.
  • the transmission parameters of the service information can be obtained according to the node capability information of the head node, the node capability information of the tail node, and the node capability information of the intermediate nodes passed from the head node to the tail node.
  • the transmission parameter of the service information may be a transmission parameter obtained by prediction, or may be a transmission parameter obtained by real-time detection from the network, which is not specifically limited in this embodiment.
  • the transmission parameters of the business information are transmission parameters obtained through prediction processing
  • a prediction model constructed by a neural network, etc. can be used to analyze the node capability information of the head node, the node capability information of the tail node, and the data generated by the head node. Contents such as node capability information of intermediate nodes passed by the tail node are predicted and processed, thereby obtaining the transmission parameters of the service information.
  • the training data used to train the network model is also the transmission parameter of the business information.
  • the training data may be acquired in various manners, which are not specifically limited in this embodiment. For example, it can be obtained by acquiring the transmission parameters of the actually transmitted service information from the network, or it can be obtained by using artificial simulation.
  • artificial simulation is used to obtain the training data, a virtual network topology can be constructed according to the actual network topology, and then the transmission of service information is simulated in the virtual network topology to obtain the training data.
  • the network model including the GCN model is trained by using the transmission parameters of the service information in the network. Since the GCN model is obtained according to the transmission path information of the service information in the network, so The GCN model can be consistent with the network topology that transmits the service information, and because the transmission parameters represent the transmission performance of the service information in the transmission path corresponding to the network, so by using the transmission parameters of the service information to include the GCN The network model of the model is trained, so that the network model including the GCN model after the training is completed can accurately and effectively realize the prediction of the transmission performance of the service information in the network.
  • the GCN model may include an input node for inputting transmission parameters, the input node corresponds to the transmission path information, and the input nodes corresponding to the transmission path information with the homologous relationship or the homoclinic relationship are connected.
  • the GCN model in this embodiment is the GCN model in the embodiment shown in FIG. 2
  • the description of the specific structure of the GCN model in this embodiment reference may be made to the GCN model shown in FIG. 3 .
  • the detailed content description in the example will not be repeated here.
  • the network model may also include a DNN model, and the input of the DNN model is connected to the output of the GCN model.
  • the network model is performed using the transmission parameters. Training can include but is not limited to the following steps:
  • Step S321 input the transmission parameters to the input node to train the GCN model to obtain transmission feature samples
  • Step S322 the transmission feature samples are input into the DNN model to train the DNN model.
  • the transmission parameters of the service information at different times can be obtained, and then, these transmission parameters are input to the input node to train the GCN model, so that the GCN model can perform these transmission parameters according to the time dimension and the space dimension.
  • Feature extraction, transmission feature samples are obtained, and then, these transmission feature samples are input into the DNN model to train the DNN model, so that the GCN model and DNN model after training can be more accurately and effectively realized the business information in the network. Transmission performance prediction.
  • An example is used to illustrate the training process of the network model using transmission parameters.
  • the transmission parameters of the service information at different times are obtained as the initial training sample data, such as time t and time t+1, etc.
  • the transmission parameters of the business information corresponding to the n input nodes at the moment are used as training sample data, and then these training sample data are input into the GCN model processing part as shown in Figure 5, so that these training sample data can be used for the GCN model.
  • the GCN model performs feature extraction on the transmission parameters according to the time dimension, which means that the GCN model performs feature extraction on the transmission parameters at different times.
  • the GCN model performs feature extraction on the transmission parameters according to the spatial dimension, which means that the GCN model performs convolution processing on the transmission parameters to obtain feature parameters with spatial characteristics.
  • inputting the transmission parameters to the input node in step S321 to train the GCN model may include, but is not limited to, the following steps:
  • the transmission parameters are input to the input node, and the GCN model is trained using the back-propagation algorithm.
  • step S322 may include, but is not limited to, the following steps:
  • the transmission feature samples are input into the DNN model, and the back-propagation algorithm is used to train the DNN model.
  • y ⁇ R n represents the transmission characteristic parameters with temporal and spatial characteristics obtained after the GCN model performs graph convolution on the transmission parameters
  • n is the number of service information transmission in the network (that is, the transmission path for transmitting service information).
  • x ⁇ R M ⁇ n ⁇ k representing n transmission parameters input to the GCN model at M times, and each transmission parameter is a k-dimensional transmission performance attribute
  • ⁇ jt ⁇ R n representing The convolution kernel parameter of the j-th dimension transmission performance attribute of n transmission parameters at the t-th time
  • x jt ⁇ R n representing the j-th dimension transmission of the n transmission parameters input to the GCN model at the t-th time performance properties.
  • an embodiment of the present application also provides an electronic device, the electronic device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor.
  • the processor and memory may be connected by a bus or otherwise.
  • the memory can be used to store non-transitory software programs and non-transitory computer-executable programs.
  • the memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one magnetic disk storage device, flash memory device, or other non-transitory solid state storage device.
  • the memory may include memory located remotely from the processor, which may be connected to the processor through a network. Examples of such networks include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.
  • the electronic device in this embodiment may be applied as a network controller in the embodiment shown in FIG. 1 or a head node for sending service information.
  • the electronic device in this embodiment is the same as that shown in FIG. 1 .
  • the network controller or the head node used for sending the service information in the exemplary embodiment belong to the same inventive concept, so these embodiments have the same implementation principle and technical effect, which will not be described in detail here.
  • the non-transitory software programs and instructions required to implement the information processing method of the above embodiment are stored in the memory, and when executed by the processor, the information processing method in the above embodiment is executed, for example, the above-described FIG. 2 is executed.
  • Method steps S110 to S120, method steps S121 to S122 in FIG. 4 are stored in the memory, and when executed by the processor, the information processing method in the above embodiment is executed, for example, the above-described FIG. 2 is executed.
  • an embodiment of the present application also provides an electronic device, the electronic device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor.
  • the processor and memory may be connected by a bus or otherwise.
  • the memory can be used to store non-transitory software programs and non-transitory computer-executable programs.
  • the memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one magnetic disk storage device, flash memory device, or other non-transitory solid state storage device.
  • the memory may include memory located remotely from the processor, which may be connected to the processor through a network. Examples of such networks include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.
  • the electronic device in this embodiment may be applied as a network controller in the embodiment shown in FIG. 1 or a head node for sending service information.
  • the electronic device in this embodiment is the same as that shown in FIG. 1 .
  • the network controller or the head node used for sending the service information in the exemplary embodiment belong to the same inventive concept, so these embodiments have the same implementation principle and technical effect, which will not be described in detail here.
  • the non-transitory software programs and instructions required to implement the method for generating a network model of the above-mentioned embodiment are stored in the memory, and when executed by the processor, the method for generating a network model in the above-mentioned embodiment is executed, for example, the above-described method is executed.
  • Method steps S210 to S220 in FIG. 6 and method steps S221 to S222 in FIG. 7 are stored in the memory, and when executed by the processor, the method for generating a network model in the above-mentioned embodiment is executed, for example, the above-described method is executed.
  • an embodiment of the present application also provides an electronic device, the electronic device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor.
  • the processor and memory may be connected by a bus or otherwise.
  • the memory can be used to store non-transitory software programs and non-transitory computer-executable programs.
  • the memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one magnetic disk storage device, flash memory device, or other non-transitory solid state storage device.
  • the memory may include memory located remotely from the processor, which may be connected to the processor through a network. Examples of such networks include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.
  • the electronic device in this embodiment may be applied as a network controller in the embodiment shown in FIG. 1 or a head node for sending service information.
  • the electronic device in this embodiment is the same as that shown in FIG. 1 .
  • the network controller or the head node used for sending the service information in the exemplary embodiment belong to the same inventive concept, so these embodiments have the same implementation principle and technical effect, which will not be described in detail here.
  • the non-transitory software programs and instructions required to realize the training method of the network model of the above-mentioned embodiment are stored in the memory, and when executed by the processor, the training method of the network model of the above-mentioned embodiment is executed, for example, the above-described method is executed.
  • Method steps S310 to S320 in FIG. 12 method steps S321 to S322 in FIG. 13 .
  • the device embodiments described above are only illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place, or may be distributed to multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution in this embodiment.
  • an embodiment of the present application also provides a computer-readable storage medium, where the computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are executed by a processor or controller, for example, by the above-mentioned Executed by a processor in the embodiment of the electronic device, the above-mentioned processor can execute the information processing method in the above-mentioned embodiment, for example, the above-described method steps S110 to S120 in FIG. 2 and method steps S121 to S121 in FIG.
  • the above-mentioned processor can be made to execute the method for generating a network model in the above-mentioned embodiment, for example, to execute the above-described method steps S210 to S220 in FIG. 6 , The method steps S221 to S222 in FIG. 7; or, executed by a processor in the above-mentioned electronic device embodiment, can cause the above-mentioned processor to execute the network model training method in the above-mentioned embodiment, for example, execute the above-described FIG. 12 method steps S310 to S320 in FIG. 13 , method steps S321 to S322 in FIG. 13 .
  • the embodiments of the present application include: acquiring transmission parameters of service information in the network, where the transmission parameters represent the transmission performance of service information in the network; and using a pre-trained network model to predict the transmission performance of the transmission parameters, wherein the network model includes a graph Convolutional network GCN model, the GCN model is obtained according to the transmission path information of service information in the network.
  • the GCN model since the GCN model is obtained according to the transmission path information of the service information in the network, the GCN model can conform to the network topology for transmitting the service information. Therefore, by using the GCN model including the GCN model The network model is used to process the transmission parameters of the service information in the network, which can accurately and effectively realize the prediction of the transmission performance of the service information in the network.
  • Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disk (DVD) or other optical disk storage, magnetic cartridges, magnetic tape, magnetic disk storage or other magnetic storage devices, or may Any other medium used to store desired information and which can be accessed by a computer.
  • communication media typically embodies computer readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and can include any information delivery media, as is well known to those of ordinary skill in the art .

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

一种信息处理方法、模型的生成及训练方法、电子设备和介质。其中，信息处理方法包括：获取网络中业务信息的传输参数，其中，传输参数表征业务信息在网络中的传输性能(S110)；利用预先训练好的网络模型对传输参数进行传输性能预测，其中，网络模型包括GCN模型，GCN模型为根据网络中业务信息的传输路径信息而得到(S120)。

Description

信息处理方法、模型的生成及训练方法、电子设备和介质
相关申请的交叉引用
本申请基于申请号为202011312980.0、申请日为2020年11月20日的中国专利申请提出，并要求该中国专利申请的优先权，该中国专利申请的全部内容在此引入本申请作为参考。
技术领域
本申请实施例涉及但不限于网络技术领域，尤其涉及一种信息处理方法、网络模型的生成方法、网络模型的训练方法、电子设备及计算机可读存储介质。
背景技术
随着网络的快速发展，网络拓扑越来越复杂，为了保证网络中业务传输的稳定性以及高质量，有必要对业务的传输性能进行预测。然而，目前对网络中业务的传输性能预测主要还是依靠运维人员的经验而进行，智能化程度较低，虽然在本领域的一些情形中提出了采用机器学习算法尝试应对业务的传输性能预测，但影响业务的传输性能预测的因素包括时间、空间、社会环境等，这些因素多且复杂，导致简单的机器学习算法在应对这种受多维因素影响的传输性能预测功能的实现时略显吃力，从而不能有效实现对网络中业务的传输性能的预测。
发明内容
以下是对本文详细描述的主题的概述。本概述并非是为了限制权利要求的保护范围。
本申请实施例提供了一种信息处理方法、网络模型的生成方法、网络模型的训练方法、电子设备及计算机可读存储介质，能够有效实现对网络中业务信息的传输性能的预测。
第一方面，本申请实施例提供了一种信息处理方法，包括：获取网络中业务信息的传输参数，其中，所述传输参数表征所述业务信息在网络中的传输性能；利用预先训练好的网络模型对所述传输参数进行传输性能预测，其中，所述网络模型包括图卷积网络(Graph Convolution Network，GCN)模型，所述GCN模型为根据网络中业务信息的传输路径信息而得到。
第二方面，本申请实施例还提供了一种网络模型的生成方法，包括：获取网络中业务信息的传输路径信息；根据所述传输路径信息生成GCN模型，其中，所述GCN模型包括用于输入所述业务信息的传输参数的输入节点，所述输入节点与所述传输路径信息相对应，所述传输参数表征所述业务信息在网络中的传输性能。
第三方面，本申请实施例还提供了一种网络模型的训练方法，包括：获取网络中业务信息的传输参数，其中，所述传输参数表征所述业务信息在网络中的传输性能；利用所述传输参数对所述网络模型进行训练，其中，所述网络模型包括GCN模型，所述GCN模型为根据网络中业务信息的传输路径信息而得到。
第四方面，本申请实施例还提供了一种电子设备，包括：存储器、处理器及存储在存储器上并可在处理器上运行的计算机程序，所述处理器执行所述计算机程序时实现如上所述第一方面的信息处理方法。
第五方面，本申请实施例还提供了一种电子设备，包括：存储器、处理器及存储在存储器上并可在处理器上运行的计算机程序，所述处理器执行所述计算机程序时实现如上所述第 二方面的网络模型的生成方法。
第六方面，本申请实施例还提供了一种电子设备，包括：存储器、处理器及存储在存储器上并可在处理器上运行的计算机程序，所述处理器执行所述计算机程序时实现如上所述第三方面的网络模型的训练方法。
第七方面，本申请实施例还提供一种计算机可读存储介质，存储有计算机可执行指令，所述计算机可执行指令用于执行如上所述第一方面的信息处理方法，或执行如上所述第二方面的网络模型的生成方法，或执行如上所述第三方面的网络模型的训练方法。
本申请的其它特征和优点将在随后的说明书中阐述，并且，部分地从说明书中变得显而易见，或者通过实施本申请而了解。本申请的目的和其他优点可通过在说明书、权利要求书以及附图中所特别指出的结构来实现和获得。
附图说明
附图用来提供对本申请技术方案的进一步理解，并且构成说明书的一部分，与本申请的实施例一起用于解释本申请的技术方案，并不构成对本申请技术方案的限制。
图1是本申请一个实施例提供的用于执行信息处理方法的网络拓扑的示意图；
图2是本申请一个实施例提供的信息处理方法的流程图；
图3是本申请一个实施例提供的GCN模型的结构示意图；
图4是本申请另一实施例提供的信息处理方法中进行传输性能预测的流程图；
图5是本申请一个示例提供的对传输参数进行传输性能预测的预测模型系统的示意图；
图6是本申请一个实施例提供的网络模型的生成方法的流程图；
图7是本申请另一实施例提供的网络模型的生成方法中生成GCN模型的流程图；
图8是本申请一个具体示例提供的业务网络虚拟拓扑的示意图；
图9是本申请一个具体示例提供的节点链路化处理示意图；
图10是本申请一个具体示例提供的网络拓扑的示意图；
图11是本申请另一具体示例提供的根据图10的网络拓扑构建而成的GCN模型的示意图；
图12是本申请一个实施例提供的网络模型的训练方法的流程图；
图13是本申请另一实施例提供的网络模型的训练方法中对网络模型进行训练的流程图。
具体实施方式
为了使本申请的目的、技术方案及优点更加清楚明白，以下结合附图及实施例，对本申请进行进一步详细说明。应当理解，此处所描述的具体实施例仅用以解释本申请，并不用于限定本申请。
需要说明的是，虽然在流程图中示出了逻辑顺序，但是在某些情况下，可以以不同于流程图中的顺序执行所示出或描述的步骤。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。
本申请提供了一种信息处理方法、网络模型的生成方法、网络模型的训练方法、电子设备及计算机可读存储介质，其中，信息处理方法包括：获取网络中业务信息的传输参数，其中，传输参数表征业务信息在网络中的传输性能；利用预先训练好的网络模型对传输参数进行传输性能预测，其中，网络模型包括GCN模型，GCN模型为根据网络中业务信息的传输路径信息而得到。由于GCN模型为根据网络中业务信息的传输路径信息而得到，因此该GCN模型能够与传输该业务信息的网络拓扑相符合，所以，通过利用包括有该GCN模型的网络模型 对网络中的业务信息的传输参数进行处理，能够较为准确有效的实现对网络中的业务信息的传输性能预测。
下面结合附图，对本申请实施例作进一步阐述。
如图1所示，图1是本申请一个实施例提供的用于执行信息处理方法的网络拓扑的示意图。在图1的示例中，该网络拓扑包括网络控制器(图中未示出)和多个节点(如图1中的节点A、节点B、......、节点S)，这些节点通过链路连接构成如图1所示的网络拓扑，并且，网络控制器与各个节点相连接。
其中，网络控制器可以获取由各个节点所上报的节点信息及业务信息等信息，从而可以根据这些节点信息形成网络拓扑模型，并可以根据该网络拓扑模型进行业务信息的传输路径的规划配置，当网络控制器完成业务信息的传输路径的规划配置而得到转发策略后，网络控制器会将转发策略下发至对应的节点，使得网络中的节点能够根据转发策略进行业务信息的传输。例如，在如图1所示的网络拓扑中，存在如下业务传输路径：从节点A传输至节点L的第一业务传输路径、从节点A传输至节点P的第二业务传输路径、从节点A传输至节点Q的第三业务传输路径、从节点B传输至节点P的第四业务传输路径、从节点B传输至节点L的第五业务传输路径、从节点B传输至节点H的第六业务传输路径、从节点L传输至节点Q的第七业务传输路径、从节点P传输至节点Q的第八业务传输路径、从节点H传输至节点Q的第九业务传输路径。
此外，网络控制器在对业务信息的传输路径进行规划配置时，还可以根据由各个节点所上传的节点信息及业务信息等信息，对业务信息进行传输性能预测，并可以对预测结果进行相应的优化处理。
另外，在分布式网络系统中，各个节点可以获取由网络中的其他节点所通告的节点信息，从而可以根据这些节点信息得到网络拓扑信息。而对于发送业务信息的首节点，可以根据该网络拓扑信息形成对应的网络拓扑模型，并且可以根据所要传输的业务信息的传输性能要求，进行业务信息的传输路径的规划配置，并得到符合要求的转发策略，接着按照该转发策略转发业务信息。
此外，发送业务信息的首节点在对业务信息的传输路径进行规划配置时，还可以对业务信息进行传输性能预测，并可以对预测结果进行相应的优化处理。
本申请实施例描述的网络拓扑以及应用场景是为了更加清楚的说明本申请实施例的技术方案，并不构成对于本申请实施例提供的技术方案的限定，本领域技术人员可知，随着网络拓扑的演变和新应用场景的出现，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。
本领域技术人员可以理解的是，图1中示出的网络拓扑结构并不构成对本申请实施例的限定，可以包括比图示更多或更少的部件，或者组合某些部件，或者不同的部件布置。
在图1所示的网络拓扑中，网络控制器或者发送业务信息的首节点可以调用其储存的信息处理程序，以执行信息处理方法。
基于上述网络拓扑的结构，提出本申请的各个实施例。
如图2所示，图2是本申请一个实施例提供的信息处理方法的流程图，该信息处理方法应用于网络控制器或者发送业务信息的首节点，该信息处理方法包括但不限于有步骤S110和步骤S120。
步骤S110，获取网络中业务信息的传输参数，其中，传输参数表征业务信息在网络中的传输性能。
在一实施例中，传输参数表征业务信息在网络中的传输性能，传输参数包括但不限于业务流量、光信噪比(Optical Signal Noise Ratio，OSNR)、误码率和出光功率等。可以理解的是，可以根据首节点的节点能力信息、尾节点的节点能力信息、由首节点至尾节点所经过的中间节点的节点能力信息等内容，获取业务信息的传输参数。此外，业务信息的传输参数可以为预测得到的传输参数，也可以为从网络中实时检测得到的传输参数，本实施例对此并不作具体限定。例如，当业务信息的传输参数为通过预测处理而得到的传输参数时，可以利用由神经网络等方式构建的预测模型等，对首节点的节点能力信息、尾节点的节点能力信息、由首节点至尾节点所经过的中间节点的节点能力信息等内容进行预测处理，从而得到业务信息的传输参数。
步骤S120，利用预先训练好的网络模型对传输参数进行传输性能预测，其中，网络模型包括GCN模型，GCN模型为根据网络中业务信息的传输路径信息而得到。
可以理解的是，由于GCN模型为根据网络中业务信息的传输路径信息而得到，因此GCN模型能够与传输该业务信息的网络拓扑相符合，所以，利用预先训练好的并且包括有该GCN模型的网络模型对业务信息的传输参数进行处理，能够较为准确有效的实现对网络中的业务信息的传输性能预测。
可以理解的是，网络模型可以为预先训练好并存储在网络控制器中或者存储在发送业务信息的首节点中，因此，在获取到网络中业务信息的传输参数后，即可利用网络模型对业务信息进行传输性能预测。由于网络模型是对业务信息的传输参数进行传输性能预测的，因此用于训练网络模型的训练数据，同样为业务信息的传输参数，可以理解的是，用于训练网络模型的训练数据，可以有多种获取方式，本实施例对此并不作具体限定。例如，可以通过从网络中获取实际传输的业务信息的传输参数而得到，也可以通过利用人工模拟的方式获取得到。当利用人工模拟的方式以得到训练数据时，可以根据实际网络拓扑构建虚拟网络拓扑，然后在虚拟网络拓扑中模拟业务信息的传输以获取得到训练数据。
通过采用包括有上述步骤S110和步骤S120的信息处理方法，利用预先训练好的包括有GCN模型的网络模型对业务信息的传输参数进行传输性能预测，由于GCN模型为根据网络中业务信息的传输路径信息而得到，因此该GCN模型能够与传输该业务信息的网络拓扑相符合，所以，通过利用包括有该GCN模型的网络模型对网络中的业务信息的传输参数进行处理，能够较为准确有效的实现对网络中的业务信息的传输性能预测。
在一实施例中，由于GCN模型为根据网络中业务信息的传输路径信息而得到，因此，如图3所示，GCN模型可以包括用于输入传输参数的输入节点，该输入节点与传输路径信息相对应，存在同源关系或者同宿关系的传输路径信息所对应的输入节点之间相连接。
在图3的示例中，GCN模型包括有输入节点AL、输入节点AP、输入节点AQ、输入节点BP、输入节点BL、输入节点BH、输入节点LQ、输入节点PQ和输入节点HQ，输入节点AL对应于如图1所示网络拓扑中的第一业务传输路径，输入节点AP对应于如图1所示网络拓扑中的第二业务传输路径，输入节点AQ对应于如图1所示网络拓扑中的第三业务传输路径，输入节点BP对应于如图1所示网络拓扑中的第四业务传输路径，输入节点BL对应于如图1所示网络拓扑中的第五业务传输路径，输入节点BH对应于如图1所示网络拓扑中的第六业务传输 路径，输入节点LQ对应于如图1所示网络拓扑中的第七业务传输路径，输入节点PQ对应于如图1所示网络拓扑中的第八业务传输路径，输入节点HQ对应于如图1所示网络拓扑中的第九业务传输路径。其中，第一业务传输路径、第二业务传输路径和第三业务传输路径具有相同的源节点(即节点A)，即存在同源关系，因此输入节点AL、输入节点AP和输入节点AQ两两连接，类似的，输入节点BP、输入节点BL和输入节点BH两两连接，输入节点AQ、输入节点LQ、输入节点PQ和输入节点HQ两两连接，输入节点AL、输入节点BL和输入节点LQ两两连接，输入节点AP、输入节点BP和输入节点PQ两两连接，输入节点BH和输入节点HQ相连接。
可以理解的是，由于传输参数表征业务信息在网络中的传输性能，即，传输参数表征业务信息在网络中对应的传输路径中进行传输时的传输性能，因此，通过采用上述根据业务信息的传输路径信息而得到的GCN模型对业务信息的传输参数进行处理，能够将基于业务信息的传输性能属性(即传输参数)转换成满足GCN模型处理的需要输入到各个输入节点的参数，从而使得包括有GCN模型的网络模型能够对网络中的业务信息进行传输性能预测。
另外，在一实施例中，网络模型还可以包括有深度神经网络(Deep Neural Network，DNN)模型，DNN模型的输入连接GCN模型的输出，在这种情况下，如图4所示，步骤S120中的利用预先训练好的网络模型对传输参数进行传输性能预测，可以包括但不限于有以下步骤：
步骤S121，利用GCN模型对传输参数进行处理，得到传输特征参数；
步骤S122，利用DNN模型对传输特征参数进行处理，得到传输性能预测结果。
在一实施例中，可以获取不同时刻下的业务信息的传输参数，而后，利用GCN模型对这些传输参数按照时间维度和空间维度进行特征提取，得到传输特征参数，接着，把这些传输特征参数作为DNN模型的输入而传输至DNN模型中，使得DNN模型能够对这些传输特征参数进行处理以得到针对业务信息的未来时刻的传输性能预测结果。
以一个示例进行说明，在如图5所示的预测模型系统中，首先分别获取各个时刻下的业务信息的传输参数，例如获取t时刻和t+1时刻等时刻下的n个输入节点所对应的业务信息的传输参数，即如图5所示的输入部分中n个输入节点的输入参数，接着，把这些输入参数输入到如图5所示的GCN模型处理部分中，使得GCN模型能够对这些输入参数进行处理，得到与下一时刻对应的传输特征参数，接着，把该传输特征参数输入到如图5所示的DNN模型处理部分中，使得DNN模型能够对该传输特征参数进行处理，得到对业务信息于下一时刻的传输性能预测结果，因此，可以利用该传输性能预测结果作为优化处理的指导信息而对业务信息进行传输性能优化处理，以改善业务信息的传输性能。
值得注意的是，利用GCN模型对传输参数按照时间维度进行特征提取，指的是利用GCN模型对不同时刻下的传输参数进行特征提取的处理。利用GCN模型对传输参数按照空间维度进行特征提取，指的是利用GCN模型对传输参数进行卷积处理以得到具有空间特征化的特征参数。
另外，在一实施例中，GCN模型可以采用公式y＝Θ*gx＝σ(α(L)x)来表征，其中，y为传输特征参数，即，y为GCN模型对传输参数进行预测处理后的输出参数；x为传输参数，即，x为输入至GCN模型的输入参数；Θ为卷积核函数；*g为采用谱方法的图卷积算子，表示x与Θ的乘积；α为卷积核参数；L为归一化的图拉普拉斯算子；σ为激活函数，例如ReLU激活函数。
可以理解的是，当获取到网络中业务信息的传输参数后，将传输参数作为GCN模型的输入参数而输入到公式y＝Θ*gx＝σ(α(L)x)中，则可以得到：
Figure PCTCN2021116968-appb-000001
其中，y∈R n，表示GCN模型对传输参数进行图卷积后得到的具有时间特征化及空间特征化的传输特征参数，n为网络中业务信息的传输数量(即传输业务信息的传输路径的数量)；x∈R M×n×k，表示输入至GCN模型的在M个时刻下的n个传输参数，并且每一个传输参数均为k维传输性能属性；α jt∈R n，表示在第t个时刻下n个传输参数的第j维传输性能属性的卷积核参数；x jt∈R n，表示输入至GCN模型的在第t个时刻下n个传输参数的第j维传输性能属性。
另外，在一实施例中，在步骤S122之后，该信息处理方法还可以包括但不限于有以下步骤：
利用传输性能预测结果对业务信息进行传输性能优化处理。
可以理解的是，由于GCN模型为根据网络中业务信息的传输路径信息而得到，因此GCN模型能够与传输业务信息的网络拓扑相符合，所以，在利用包括有GCN模型的网络模型对网络中的业务信息的传输参数进行传输性能预测而得到传输性能预测结果后，可以利用该传输性能预测结果作为优化处理的指导信息而对业务信息进行传输性能优化处理，以改善业务信息的传输性能。其中，利用传输性能预测结果对业务信息进行传输性能优化处理，可以有不同的处理方式，本实施例对此并不作具体限定，例如，可以根据传输性能预测结果调整业务信息在相应的传输路径中的性能参数以达到优化传输性能的目的，也可以根据传输性能预测结果改变业务信息在网络中的传输路径以达到优化传输性能的目的。
另外，如图6所示，图6是本申请另一个实施例提供的网络模型的生成方法的流程图，该网络模型的生成方法应用于网络控制器或者发送业务信息的首节点，该网络模型的生成方法包括但不限于有以下步骤：
步骤S210，获取网络中业务信息的传输路径信息；
步骤S220，根据传输路径信息生成GCN模型，其中，GCN模型包括用于输入业务信息的传输参数的输入节点，输入节点与传输路径信息相对应，传输参数表征业务信息在网络中的传输性能。
可以理解的是，本实施例中根据传输路径信息而生成的GCN模型，即为如图2所示实施例中的用于对网络中业务信息的传输参数进行传输性能预测的GCN模型。
可以理解的是，网络中业务信息的传输路径信息可以包括有首节点信息和尾节点信息，其中，首节点信息包括发送业务信息的首节点的地址信息，尾节点信息包括接收业务信息的尾节点的地址信息。当首节点需要向尾节点发送业务信息时，可以从首节点中获取得到首节点的地址信息和尾节点的地址信息，从而能够获取到该业务信息的传输路径信息。
在一实施例中，传输参数表征业务信息在网络中的传输性能，传输参数包括但不限于业务流量、OSNR、误码率和出光功率等。可以理解的是，可以根据首节点的节点能力信息、尾节点的节点能力信息、由首节点至尾节点所经过的中间节点的节点能力信息等内容，获取业务信息的传输参数。此外，业务信息的传输参数可以为预测得到的传输参数，也可以为从网 络中实时检测得到的传输参数，本实施例对此并不作具体限定。例如，当业务信息的传输参数为通过预测处理而得到的传输参数时，可以利用由神经网络等方式构建的预测模型等，对首节点的节点能力信息、尾节点的节点能力信息、由首节点至尾节点所经过的中间节点的节点能力信息等内容进行预测处理，从而得到业务信息的传输参数。
当获取到网络中业务信息的传输路径信息后，即可根据该传输路径信息生成包括有用于输入业务信息的传输参数的输入节点的GCN模型，该GCN模型的输入节点与传输路径信息相对应，例如，如图3中所示的输入节点AL与如图1中所示的第一业务传输路径相对应。由于传输参数表征业务信息在网络中的传输性能，而GCN模型为根据网络中业务信息的传输路径信息而得到，因此，利用传输参数作为GCN模型的输入节点的输入参数，能够较为准确有效的实现对网络中的业务信息的传输性能预测。
另外，在一实施例中，如图7所示，步骤S220中的根据传输路径信息生成GCN模型，可以包括但不限于有以下步骤：
步骤S221，根据首节点信息和尾节点信息建立输入节点；
步骤S222，将存在同源关系或者同宿关系的传输路径信息所对应的输入节点相连接，得到GCN模型。
可以理解的是，通过根据传输业务信息的首节点信息和尾节点信息而建立GCN模型的输入节点，并且将存在同源关系或者同宿关系的传输路径信息所对应的输入节点相连接，可以使得由此得到的GCN模型能够对业务信息在对应的传输路径中的传输参数进行预测处理，能够较为准确有效的实现对网络中的业务信息的传输性能预测。
下面以具体的示例对GCN模型的生成进行详细的说明。
在如图1所示的网络拓扑中，针对相邻节点间单条链路的情况，根据每条业务路径的首节点和尾节点，忽略中间节点的情况，抽象成如图8所示的业务网络虚拟拓扑，其中，该业务网络虚拟拓扑中仅包括每条业务路径的首节点和尾节点，并且对于每条业务路径，其首节点和尾节点直接连接以形成业务虚拟链路。在将如图1所示的网络拓扑抽象成如图8所示的业务网络虚拟拓扑后，根据如图8所示的业务网络虚拟拓扑中各个节点的出度数，以及业务虚拟链路共享同一末端节点的两两依赖关系(如同源关系或者同宿关系)，针对每一个节点，均可以抽象转换成
Figure PCTCN2021116968-appb-000002
条虚拟路径，其中z表示该节点的出度数。以图8中的节点L为例，包括有业务虚拟链路AL、业务虚拟链路BL和业务虚拟链路LQ共享节点L作为末端节点，因此，参照如图9所示的节点链路化处理示意图，节点L可以被抽象转换成3条虚拟路径。此外，在将如图1所示的网络拓扑抽象成如图8所示的业务网络虚拟拓扑后，将业务网络虚拟拓扑中的每条业务虚拟链路抽象转换成GCN模型的输入节点，即根据首节点信息和尾节点信息建立输入节点，例如，将从节点A至节点L的业务虚拟链路抽象转换成输入节点AL。在将如图8所示的业务网络虚拟拓扑中的全部业务虚拟链路均抽象转换成输入节点后，根据如图9所示的链路化处理，将存在同源关系或者同宿关系的业务虚拟链路所对应的输入节点相连接，即可得到如图3所示的利用一个虚拟节点(即输入节点)表征一条业务路径的GCN模型。由于GCN模型为根据网络中业务信息的传输路径信息而得到，因此GCN模型能够与传输业务信息的网络拓扑相符合，所以，通过利用GCN模型对网络中的业务信息的传输参数进行相关的预测处理，能够较为准确有效的实现对网络中的业务信息的传输性能预测。
值得注意的是，针对相邻节点间多条链路的情况，可以通过增加虚拟假节点的方式，把 相邻节点间多条链路的网络拓扑形式转换成相邻节点间单条链路的网络拓扑形式，进而能够采用上述方式构建GCN模型。例如，假设在如图10所示的网络拓扑形式中，节点A和节点M之间存在3条业务虚拟链路，那么，可以通过增加虚拟假节点的方式，将这3条业务虚拟链路抽象转换成输入节点AM1、输入节点AM2和输入节点AM3，并结合根据节点A至节点Q的业务虚拟链路所抽象转换成的输入节点AQ、根据节点Q至节点R的业务虚拟链路所抽象转换成的输入节点QR、根据节点M至节点R的业务虚拟链路所抽象转换成的输入节点MR，构建成如图11所示的GCN模型。
此外，值得注意的是，如果业务路径需要经过中间节点作为电中继，那么整条业务路径可根据首节点、尾节点和电中继节点，被划分成多个业务路径段，并且每个业务路径段均被认为是一条独立的业务路径而进行虚拟化处理。
在一实施例中，根据网络中业务信息的传输路径信息而生成的GCN模型，可以采用公式y＝Θ*gx＝σ(α(L)x)来表征，其中，y为该GCN模型的输出参数；x为输入到该GCN模型的输入参数；Θ为卷积核函数；*g为采用谱方法的图卷积算子，表示x与Θ的乘积；α为卷积核参数；L为归一化的图拉普拉斯算子；σ为激活函数，例如ReLU激活函数。
可以理解的是，当获取到网络中业务信息的传输参数后，可以将该传输参数作为GCN模型的输入参数而输入到公式y＝Θ*gx＝σ(α(L)x)中，则可以得到：
Figure PCTCN2021116968-appb-000003
其中，y∈R n，表示GCN模型对传输参数进行图卷积后得到的具有时间特征化及空间特征化的传输特征参数，n为网络中业务信息的传输数量(即传输业务信息的传输路径的数量)；x∈R M×n×k，表示输入至GCN模型的在M个时刻下的n个传输参数，并且每一个传输参数均为k维传输性能属性；α jt∈R n，表示在第t个时刻下n个传输参数的第j维传输性能属性的卷积核参数；x jt∈R n，表示输入至GCN模型的在第t个时刻下n个传输参数的第j维传输性能属性。
另外，如图12所示，图12是本申请另一个实施例提供的网络模型的训练方法的流程图，该网络模型的训练方法应用于网络控制器或者发送业务信息的首节点，该网络模型的训练方法包括但不限于有以下步骤：
步骤S310，获取网络中业务信息的传输参数，其中，传输参数表征业务信息在网络中的传输性能；
步骤S320，利用传输参数对网络模型进行训练，其中，网络模型包括GCN模型，GCN模型为根据网络中业务信息的传输路径信息而得到。
可以理解的是，本实施例中利用网络中业务信息的传输参数来进行训练的GCN模型，即为如图2所示实施例中的预先训练好的GCN模型，也为如图6所示实施例中根据传输路径信息而生成的GCN模型。通过利用如图6所示实施例中的生成方法得到GCN模型，而后利用本实施例中的训练方法对该GCN模型进行训练，当完成训练后，即可得到如图2所示实施例中的预先训练好的GCN模型。
可以理解的是，网络中业务信息的传输路径信息可以包括有首节点信息和尾节点信息，其中，首节点信息包括发送业务信息的首节点的地址信息，尾节点信息包括接收业务信息的 尾节点的地址信息。当首节点需要向尾节点发送业务信息时，可以从首节点中获取得到首节点的地址信息和尾节点的地址信息，从而能够获取到该业务信息的传输路径信息。
在一实施例中，传输参数表征业务信息在网络中的传输性能，传输参数包括但不限于业务流量、OSNR、误码率和出光功率等。可以理解的是，可以根据首节点的节点能力信息、尾节点的节点能力信息、由首节点至尾节点所经过的中间节点的节点能力信息等内容，获取业务信息的传输参数。此外，业务信息的传输参数可以为预测得到的传输参数，也可以为从网络中实时检测得到的传输参数，本实施例对此并不作具体限定。例如，当业务信息的传输参数为通过预测处理而得到的传输参数时，可以利用由神经网络等方式构建的预测模型等，对首节点的节点能力信息、尾节点的节点能力信息、由首节点至尾节点所经过的中间节点的节点能力信息等内容进行预测处理，从而得到业务信息的传输参数。
由于包括有GCN模型的网络模型是对业务信息的传输参数进行传输性能预测的，因此用于训练网络模型的训练数据，同样为业务信息的传输参数，可以理解的是，用于训练网络模型的训练数据，可以有多种获取方式，本实施例对此并不作具体限定。例如，可以通过从网络中获取实际传输的业务信息的传输参数而得到，也可以通过利用人工模拟的方式获取得到。当利用人工模拟的方式以得到训练数据时，可以根据实际网络拓扑构建虚拟网络拓扑，然后在虚拟网络拓扑中模拟业务信息的传输以获取得到训练数据。
通过采用包括有上述步骤S310和步骤S320的训练方法，利用网络中业务信息的传输参数对包括有GCN模型的网络模型进行训练，由于GCN模型为根据网络中业务信息的传输路径信息而得到，因此GCN模型能够与传输该业务信息的网络拓扑相符合，而由于传输参数表征业务信息在网络中对应的传输路径中进行传输时的传输性能，所以，通过利用业务信息的传输参数对包括有该GCN模型的网络模型进行训练，能够使得完成训练后的包括有该GCN模型的网络模型能够较为准确有效的实现对网络中的业务信息的传输性能预测。
在一实施例中，GCN模型可以包括用于输入传输参数的输入节点，该输入节点与传输路径信息相对应，存在同源关系或者同宿关系的传输路径信息所对应的输入节点之间相连接。
需要说明的是，由于本实施例中的GCN模型即为如图2所示实施例中的GCN模型，因此，对于本实施例中的GCN模型的具体结构的说明，可以参照如图3所示示例中的详细的内容描述，此处不再赘述。
另外，在一实施例中，网络模型还可以包括有DNN模型，DNN模型的输入连接GCN模型的输出，在这种情况下，如图13所示，步骤S320中的利用传输参数对网络模型进行训练，可以包括但不限于有以下步骤：
步骤S321，将传输参数输入至输入节点以对GCN模型进行训练，得到传输特征样本；
步骤S322，将传输特征样本输入至DNN模型以对DNN模型进行训练。
在一实施例中，可以获取不同时刻下的业务信息的传输参数，而后，利用这些传输参数输入至输入节点以对GCN模型进行训练，使得GCN模型能够对这些传输参数按照时间维度和空间维度进行特征提取，得到传输特征样本，接着，把这些传输特征样本输入至DNN模型以对DNN模型进行训练，从而使得完成训练后的GCN模型和DNN模型能够较为准确有效的实现对网络中的业务信息的传输性能预测。
以一个示例说明利用传输参数对网络模型进行训练过程，如图5所示，首先分别获取不同时刻下的业务信息的传输参数以作为初始的训练样本数据，例如获取t时刻和t+1时刻等 时刻下的n个输入节点所对应的业务信息的传输参数作为训练样本数据，接着，把这些训练样本数据输入到如图5所示的GCN模型处理部分中，使得这些训练样本数据能够对GCN模型进行训练，得到传输特征样本，接着，把该传输特征样本输入到如图5所示的DNN模型处理部分中，使得该传输特征样本能够对DNN模型进行训练，从而使得完成训练后的包括有该GCN模型的网络模型能够较为准确有效的实现对网络中的业务信息的传输性能预测。
值得注意的是，GCN模型对传输参数按照时间维度进行特征提取，指的是GCN模型对不同时刻下的传输参数进行特征提取的处理。GCN模型对传输参数按照空间维度进行特征提取，指的是GCN模型对传输参数进行卷积处理以得到具有空间特征化的特征参数。
另外，在一实施例中，步骤S321中的将传输参数输入至输入节点以对GCN模型进行训练，可以包括但不限于有以下步骤：
将传输参数输入至输入节点，并采用反向传播算法对GCN模型进行训练。
可以理解的是，采用反向传播算法对GCN模型进行训练，能够实现通过监督学习的形式对GCN模型进行训练，可以使得完成训练后的GCN模型能够更加准确的对网络中业务信息的传输参数进行传输性能预测，从而使得后续的信息处理操作能够更为准确有效的实现对网络中的业务信息的传输性能预测。
另外，在一实施例中，步骤S322可以包括但不限于有以下步骤：
将传输特征样本输入至DNN模型，并采用反向传播算法对DNN模型进行训练。
可以理解的是，采用反向传播算法对DNN模型进行训练，能够实现通过监督学习的形式对DNN模型进行训练，可以使得完成训练后的DNN模型能够更加准确的对网络中业务信息的传输参数进行传输性能预测，从而使得后续的信息处理操作能够更为准确有效的实现对网络中的业务信息的传输性能预测。
在一实施例中，根据网络中业务信息的传输路径信息而得到的GCN模型，可以采用公式y＝Θ*gx＝σ(α(L)x)来表征，其中，y为该GCN模型的输出参数；x为输入到该GCN模型的输入参数；Θ为卷积核函数；*g为采用谱方法的图卷积算子，表示x与Θ的乘积；α为卷积核参数；L为归一化的图拉普拉斯算子；σ为激活函数，例如ReLU激活函数。
可以理解的是，当获取到网络中业务信息的传输参数后，可以将该传输参数作为训练样本数据而输入到公式y＝Θ*gx＝σ(α(L)x)中，以对GCN模型进行训练。需要说明的是，当将传输参数作为训练样本数据而输入到公式y＝Θ*gx＝σ(α(L)x)后，可以得到：
Figure PCTCN2021116968-appb-000004
其中，y∈R n，表示GCN模型对传输参数进行图卷积后得到的具有时间特征化及空间特征化的传输特征参数，n为网络中业务信息的传输数量(即传输业务信息的传输路径的数量)；x∈R M×n×k，表示输入至GCN模型的在M个时刻下的n个传输参数，并且每一个传输参数均为k维传输性能属性；α jt∈R n，表示在第t个时刻下n个传输参数的第j维传输性能属性的卷积核参数；x jt∈R n，表示输入至GCN模型的在第t个时刻下n个传输参数的第j维传输性能属性。
另外，本申请的一个实施例还提供了一种电子设备，该电子设备包括：存储器、处理器及存储在存储器上并可在处理器上运行的计算机程序。
处理器和存储器可以通过总线或者其他方式连接。
存储器作为一种非暂态计算机可读存储介质，可用于存储非暂态软件程序以及非暂态性计算机可执行程序。此外，存储器可以包括高速随机存取存储器，还可以包括非暂态存储器，例如至少一个磁盘存储器件、闪存器件、或其他非暂态固态存储器件。在一些实施方式中，存储器可包括相对于处理器远程设置的存储器，这些远程存储器可以通过网络连接至该处理器。上述网络的实例包括但不限于互联网、企业内部网、局域网、移动通信网及其组合。
需要说明的是，本实施例中的电子设备，可以应用为如图1所示实施例中的网络控制器或者用于发送业务信息的首节点，本实施例中的电子设备和如图1所示实施例中的网络控制器或者用于发送业务信息的首节点属于相同的发明构思，因此这些实施例具有相同的实现原理以及技术效果，此处不再详述。
实现上述实施例的信息处理方法所需的非暂态软件程序以及指令存储在存储器中，当被处理器执行时，执行上述实施例中的信息处理方法，例如，执行以上描述的图2中的方法步骤S110至S120、图4中的方法步骤S121至S122。
另外，本申请的一个实施例还提供了一种电子设备，该电子设备包括：存储器、处理器及存储在存储器上并可在处理器上运行的计算机程序。
处理器和存储器可以通过总线或者其他方式连接。
存储器作为一种非暂态计算机可读存储介质，可用于存储非暂态软件程序以及非暂态性计算机可执行程序。此外，存储器可以包括高速随机存取存储器，还可以包括非暂态存储器，例如至少一个磁盘存储器件、闪存器件、或其他非暂态固态存储器件。在一些实施方式中，存储器可包括相对于处理器远程设置的存储器，这些远程存储器可以通过网络连接至该处理器。上述网络的实例包括但不限于互联网、企业内部网、局域网、移动通信网及其组合。
需要说明的是，本实施例中的电子设备，可以应用为如图1所示实施例中的网络控制器或者用于发送业务信息的首节点，本实施例中的电子设备和如图1所示实施例中的网络控制器或者用于发送业务信息的首节点属于相同的发明构思，因此这些实施例具有相同的实现原理以及技术效果，此处不再详述。
实现上述实施例的网络模型的生成方法所需的非暂态软件程序以及指令存储在存储器中，当被处理器执行时，执行上述实施例中的网络模型的生成方法，例如，执行以上描述的图6中的方法步骤S210至S220、图7中的方法步骤S221至S222。
另外，本申请的一个实施例还提供了一种电子设备，该电子设备包括：存储器、处理器及存储在存储器上并可在处理器上运行的计算机程序。
处理器和存储器可以通过总线或者其他方式连接。
存储器作为一种非暂态计算机可读存储介质，可用于存储非暂态软件程序以及非暂态性计算机可执行程序。此外，存储器可以包括高速随机存取存储器，还可以包括非暂态存储器，例如至少一个磁盘存储器件、闪存器件、或其他非暂态固态存储器件。在一些实施方式中，存储器可包括相对于处理器远程设置的存储器，这些远程存储器可以通过网络连接至该处理器。上述网络的实例包括但不限于互联网、企业内部网、局域网、移动通信网及其组合。
需要说明的是，本实施例中的电子设备，可以应用为如图1所示实施例中的网络控制器或者用于发送业务信息的首节点，本实施例中的电子设备和如图1所示实施例中的网络控制器或者用于发送业务信息的首节点属于相同的发明构思，因此这些实施例具有相同的实现原 理以及技术效果，此处不再详述。
实现上述实施例的网络模型的训练方法所需的非暂态软件程序以及指令存储在存储器中，当被处理器执行时，执行上述实施例中的网络模型的训练方法，例如，执行以上描述的图12中的方法步骤S310至S320、图13中的方法步骤S321至S322。
以上所描述的设备实施例仅仅是示意性的，其中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，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。
此外，本申请的一个实施例还提供了一种计算机可读存储介质，该计算机可读存储介质存储有计算机可执行指令，该计算机可执行指令被一个处理器或控制器执行，例如，被上述电子设备实施例中的一个处理器执行，可使得上述处理器执行上述实施例中的信息处理方法，例如，执行以上描述的图2中的方法步骤S110至S120、图4中的方法步骤S121至S122；或者，被上述电子设备实施例中的一个处理器执行，可使得上述处理器执行上述实施例中的网络模型的生成方法，例如，执行以上描述的图6中的方法步骤S210至S220、图7中的方法步骤S221至S222；或者，被上述电子设备实施例中的一个处理器执行，可使得上述处理器执行上述实施例中的网络模型的训练方法，例如，执行以上描述的图12中的方法步骤S310至S320、图13中的方法步骤S321至S322。
本申请实施例包括：获取网络中业务信息的传输参数，其中，传输参数表征业务信息在网络中的传输性能；利用预先训练好的网络模型对传输参数进行传输性能预测，其中，网络模型包括图卷积网络GCN模型，GCN模型为根据网络中业务信息的传输路径信息而得到。根据本申请实施例提供的方案，由于GCN模型为根据网络中业务信息的传输路径信息而得到，因此该GCN模型能够与传输该业务信息的网络拓扑相符合，所以，通过利用包括有该GCN模型的网络模型对网络中的业务信息的传输参数进行处理，能够较为准确有效的实现对网络中的业务信息的传输性能预测。
本领域普通技术人员可以理解，上文中所公开方法中的全部或某些步骤、系统可以被实施为软件、固件、硬件及其适当的组合。某些物理组件或所有物理组件可以被实施为由处理器，如中央处理器、数字信号处理器或微处理器执行的软件，或者被实施为硬件，或者被实施为集成电路，如专用集成电路。这样的软件可以分布在计算机可读介质上，计算机可读介质可以包括计算机存储介质(或非暂时性介质)和通信介质(或暂时性介质)。如本领域普通技术人员公知的，术语计算机存储介质包括在用于存储信息(诸如计算机可读指令、数据结构、程序模块或其他数据)的任何方法或技术中实施的易失性和非易失性、可移除和不可移除介质。计算机存储介质包括但不限于RAM、ROM、EEPROM、闪存或其他存储器技术、CD-ROM、数字多功能盘(DVD)或其他光盘存储、磁盒、磁带、磁盘存储或其他磁存储装置、或者可以用于存储期望的信息并且可以被计算机访问的任何其他的介质。此外，本领域普通技术人员公知的是，通信介质通常包含计算机可读指令、数据结构、程序模块或者诸如载波或其他传输机制之类的调制数据信号中的其他数据，并且可包括任何信息递送介质。
以上是对本申请的一些实施进行了具体说明，但本申请并不局限于上述实施方式，熟悉本领域的技术人员在不违背本申请范围的前提下还可作出种种的等同变形或替换，这些等同的变形或替换均包含在本申请权利要求所限定的范围内。

Claims (21)

  1. 一种信息处理方法，包括：
    获取网络中业务信息的传输参数，其中，所述传输参数表征所述业务信息在网络中的传输性能；
    利用预先训练好的网络模型对所述传输参数进行传输性能预测，其中，所述网络模型包括图卷积网络GCN模型，所述GCN模型为根据网络中业务信息的传输路径信息而得到。
  2. 根据权利要求1所述的方法，其中，所述GCN模型包括用于输入所述传输参数的输入节点，所述输入节点与所述传输路径信息相对应，存在同源关系或者同宿关系的传输路径信息所对应的所述输入节点之间相连接。
  3. 根据权利要求1所述的方法，其中，所述网络模型还包括深度神经网络DNN模型，所述利用预先训练好的网络模型对所述传输参数进行传输性能预测，包括：
    利用所述GCN模型对所述传输参数进行处理，得到传输特征参数；
    利用所述DNN模型对所述传输特征参数进行处理，得到传输性能预测结果。
  4. 根据权利要求3所述的方法，其中，所述利用所述GCN模型对所述传输参数进行处理，得到传输特征参数，包括：
    利用如下公式对所述传输参数进行处理以得到传输特征参数：
    y＝Θ*gx＝σ(α(L)x)
    其中，所述y为所述传输特征参数；所述x为所述传输参数；所述Θ为卷积核函数；所述*g为采用谱方法的图卷积算子，表示所述x与所述Θ的乘积；所述α为卷积核参数；所述L为归一化的图拉普拉斯算子；所述σ为激活函数。
  5. 根据权利要求3所述的方法，其中，在得到传输性能预测结果之后，所述方法还包括：
    利用所述传输性能预测结果对所述业务信息进行传输性能优化处理。
  6. 根据权利要求1至5任意一项所述的方法，其中，所述传输参数包括业务流量、光信噪比OSNR、误码率和出光功率中的至少一个。
  7. 一种网络模型的生成方法，包括：
    获取网络中业务信息的传输路径信息；
    根据所述传输路径信息生成GCN模型，其中，所述GCN模型包括用于输入所述业务信息的传输参数的输入节点，所述输入节点与所述传输路径信息相对应，所述传输参数表征所述业务信息在网络中的传输性能。
  8. 根据权利要求7所述的方法，其中，所述传输路径信息包括首节点信息和尾节点信息，所述根据所述传输路径信息生成GCN模型，包括：
    根据所述首节点信息和所述尾节点信息建立所述输入节点；
    将存在同源关系或者同宿关系的传输路径信息所对应的所述输入节点相连接，得到GCN模型。
  9. 根据权利要求7或8所述的方法，其中，所述GCN模型采用如下公式表征：
    y＝Θ*gx＝σ(α(L)x)
    其中，所述y为所述GCN模型的输出参数；所述x为输入到所述GCN模型的输入参数；所述Θ为卷积核函数；所述*g为采用谱方法的图卷积算子，表示所述x与所述Θ的乘积；所 述α为卷积核参数；所述L为归一化的图拉普拉斯算子；所述σ为激活函数。
  10. 根据权利要求7或8所述的方法，其中，所述传输参数包括业务流量、OSNR、误码率和出光功率中的至少一个。
  11. 一种网络模型的训练方法，包括：
    获取网络中业务信息的传输参数，其中，所述传输参数表征所述业务信息在网络中的传输性能；
    利用所述传输参数对所述网络模型进行训练，其中，所述网络模型包括GCN模型，所述GCN模型为根据网络中业务信息的传输路径信息而得到。
  12. 根据权利要求11所述的方法，其中，所述GCN模型包括用于输入所述传输参数的输入节点，所述输入节点与所述传输路径信息相对应，存在同源关系或者同宿关系的传输路径信息所对应的所述输入节点之间相连接。
  13. 根据权利要求12所述的方法，其中，所述网络模型还包括DNN模型，所述利用所述传输参数对所述网络模型进行训练，包括：
    将所述传输参数输入至所述输入节点以对所述GCN模型进行训练，得到传输特征样本；
    将所述传输特征样本输入至所述DNN模型以对所述DNN模型进行训练。
  14. 根据权利要求13所述的方法，其中，所述将所述传输参数输入至所述输入节点以对所述GCN模型进行训练，包括：
    将所述传输参数输入至所述输入节点，并采用反向传播算法对所述GCN模型进行训练。
  15. 根据权利要求13所述的方法，其中，所述将所述传输特征样本输入至所述DNN模型以对所述DNN模型进行训练，包括：
    将所述传输特征样本输入至所述DNN模型，并采用反向传播算法对所述DNN模型进行训练。
  16. 根据权利要求11至15任意一项所述的方法，其中，所述GCN模型采用如下公式表征：
    y＝Θ*gx＝σ(α(L)x)
    其中，所述y为所述GCN模型的输出参数；所述x为输入到所述GCN模型的输入参数；所述Θ为卷积核函数；所述*g为采用谱方法的图卷积算子，表示所述x与所述Θ的乘积；所述α为卷积核参数；所述L为归一化的图拉普拉斯算子；所述σ为激活函数。
  17. 根据权利要求11至15任意一项所述的方法，其中，所述传输参数包括业务流量、OSNR、误码率和出光功率中的至少一个。
  18. 一种电子设备，包括：存储器、处理器及存储在存储器上并可在处理器上运行的计算机程序，其中，所述处理器执行所述计算机程序时实现如权利要求1至6中任意一项所述的信息处理方法。
  19. 一种电子设备，包括：存储器、处理器及存储在存储器上并可在处理器上运行的计算机程序，其中，所述处理器执行所述计算机程序时实现如权利要求7至10中任意一项所述的生成方法。
  20. 一种电子设备，包括：存储器、处理器及存储在存储器上并可在处理器上运行的计算机程序，其中，所述处理器执行所述计算机程序时实现如权利要求11至17中任意一项所述的训练方法。
  21. 一种计算机可读存储介质，存储有计算机可执行指令，其中，所述计算机可执行指令用于执行权利要求1至6中任意一项所述的信息处理方法，或执行权利要求7至10中任意一项所述的生成方法，或执行权利要求11至17中任意一项所述的训练方法。
PCT/CN2021/116968 2020-11-20 2021-09-07 信息处理方法、模型的生成及训练方法、电子设备和介质 WO2022105374A1 (zh)

Priority Applications (3)

Application Number Priority Date Filing Date Title
EP21893525.2A EP4243360A4 (en) 2020-11-20 2021-09-07 INFORMATION PROCESSING METHOD, MODULE GENERATION AND DRIVE METHOD, ELECTRONIC DEVICE AND MEDIUM
JP2023528716A JP7544976B2 (ja) 2020-11-20 2021-09-07 情報処理方法、モデルの生成及びトレーニング方法、電子機器並びに媒体
US18/253,130 US20240015077A1 (en) 2020-11-20 2021-09-07 Information processing method, method for generating and training module, electronic device, and medium

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202011312980.0 2020-11-20
CN202011312980.0A CN114531363A (zh) 2020-11-20 2020-11-20 信息处理方法、模型的生成及训练方法、电子设备和介质

Publications (1)

Publication Number Publication Date
WO2022105374A1 true WO2022105374A1 (zh) 2022-05-27

Family

ID=81619146

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/116968 WO2022105374A1 (zh) 2020-11-20 2021-09-07 信息处理方法、模型的生成及训练方法、电子设备和介质

Country Status (5)

Country Link
US (1) US20240015077A1 (zh)
EP (1) EP4243360A4 (zh)
JP (1) JP7544976B2 (zh)
CN (1) CN114531363A (zh)
WO (1) WO2022105374A1 (zh)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109740785A (zh) * 2018-10-22 2019-05-10 北京师范大学 基于图卷积神经网络的节点状态预测的方法
CN110995520A (zh) * 2020-02-28 2020-04-10 清华大学 网络流量预测方法、装置、计算机设备及可读存储介质
US20200151288A1 (en) * 2018-11-09 2020-05-14 Nvidia Corp. Deep Learning Testability Analysis with Graph Convolutional Networks
CN111462088A (zh) * 2020-04-01 2020-07-28 深圳前海微众银行股份有限公司 基于图卷积神经网络的数据处理方法、装置、设备及介质

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP1056297A3 (en) 1999-05-24 2002-10-02 Agilent Technologies, Inc. (a Delaware corporation) Multimedia decoder with error detection
US11461635B2 (en) * 2017-10-09 2022-10-04 Nec Corporation Neural network transfer learning for quality of transmission prediction
EP3794511A1 (en) * 2018-05-18 2021-03-24 BenevolentAI Technology Limited Graph neutral networks with attention

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109740785A (zh) * 2018-10-22 2019-05-10 北京师范大学 基于图卷积神经网络的节点状态预测的方法
US20200151288A1 (en) * 2018-11-09 2020-05-14 Nvidia Corp. Deep Learning Testability Analysis with Graph Convolutional Networks
CN110995520A (zh) * 2020-02-28 2020-04-10 清华大学 网络流量预测方法、装置、计算机设备及可读存储介质
CN111462088A (zh) * 2020-04-01 2020-07-28 深圳前海微众银行股份有限公司 基于图卷积神经网络的数据处理方法、装置、设备及介质

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See also references of EP4243360A4 *

Also Published As

Publication number Publication date
CN114531363A (zh) 2022-05-24
EP4243360A1 (en) 2023-09-13
JP2023548943A (ja) 2023-11-21
EP4243360A4 (en) 2024-04-10
US20240015077A1 (en) 2024-01-11
JP7544976B2 (ja) 2024-09-03

Similar Documents

Publication Publication Date Title
Dao et al. Multi-tier multi-access edge computing: The role for the fourth industrial revolution
US20220114475A1 (en) Methods and systems for decentralized federated learning
Zhou et al. [Retracted] Machine Learning‐Based Offloading Strategy for Lightweight User Mobile Edge Computing Tasks
US9652720B2 (en) Triggering on-the-fly requests for supervised learning of learning machines
Dai et al. Routing optimization meets Machine Intelligence: A perspective for the future network
El Rajab et al. Zero-touch networks: Towards next-generation network automation
US20220172054A1 (en) Intermediate network node and method performed therein for handling data of communication networks
Mertens et al. i-WSN League: Clustered Distributed Learning in Wireless Sensor Networks
CN114070775A (zh) 面向5g智能网联系统的区块链网络切片安全智能优化方法
CN111538772B (zh) 数据的交换处理方法、装置、电子设备及存储介质
WO2022105374A1 (zh) 信息处理方法、模型的生成及训练方法、电子设备和介质
Kurniawan et al. Mobile computing and communications-driven fog-assisted disaster evacuation techniques for context-aware guidance support: A survey
US20230409983A1 (en) Customizable federated learning
US20230132213A1 (en) Managing bias in federated learning
Ahmed et al. Collaborative flow-identification mechanism for software-defined internet of things
CN116155805A (zh) 分布式智能路由方法、系统、电子设备和存储介质
CN115426635A (zh) 一种不可靠传输场景下无人机通信网络推断方法及系统
CN115499365A (zh) 路由优化方法、装置、设备及介质
Anand Towards evolution of M2M into Internet of Things for analytics
Deb et al. Loop-the-loops: Fragmented learning over networks for constrained IoT devices
CN106487694A (zh) 一种数据流处理方法和装置
WO2023168976A1 (zh) 光传送网性能预测方法、系统、电子设备及存储介质
WO2020261285A1 (en) First node, and method performed thereby, for handling a problem in a communications network
CN104579971B (zh) 用于软件定义的网络体系结构中移交的路由的方法和设备
US11716151B2 (en) Routing methods for quantum communication paths across a mesh quantum network

Legal Events

Date Code Title Description
WWE Wipo information: entry into national phase

Ref document number: 2023528716

Country of ref document: JP

WWE Wipo information: entry into national phase

Ref document number: 18253130

Country of ref document: US

ENP Entry into the national phase

Ref document number: 2021893525

Country of ref document: EP

Effective date: 20230607

NENP Non-entry into the national phase

Ref country code: DE