CN116436828A - Network traffic prediction and model training method, device, equipment and storage medium - Google Patents

Network traffic prediction and model training method, device, equipment and storage medium Download PDF

Info

Publication number
CN116436828A
CN116436828A CN202310228544.2A CN202310228544A CN116436828A CN 116436828 A CN116436828 A CN 116436828A CN 202310228544 A CN202310228544 A CN 202310228544A CN 116436828 A CN116436828 A CN 116436828A
Authority
CN
China
Prior art keywords
flow data
network
fusion
nodes
data
Prior art date
Legal status (The legal status 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 status listed.)
Pending
Application number
CN202310228544.2A
Other languages
Chinese (zh)
Inventor
唐瑜
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing Baidu Netcom Science and Technology Co Ltd
Original Assignee
Beijing Baidu Netcom Science and Technology Co Ltd
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 Beijing Baidu Netcom Science and Technology Co Ltd filed Critical Beijing Baidu Netcom Science and Technology Co Ltd
Priority to CN202310228544.2A priority Critical patent/CN116436828A/en
Publication of CN116436828A publication Critical patent/CN116436828A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • 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
    • G06N3/042—Knowledge-based neural networks; Logical representations of neural 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
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
    • 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
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00—Computing arrangements using knowledge-based models
    • G06N5/04—Inference or reasoning models
    • 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
    • H04L41/14—Network analysis or design
    • H04L41/145—Network analysis or design involving simulating, designing, planning or modelling of a network
    • 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
    • H04L43/00—Arrangements for monitoring or testing data switching networks
    • H04L43/08—Monitoring or testing based on specific metrics, e.g. QoS, energy consumption or environmental parameters
    • H04L43/0876—Network utilisation, e.g. volume of load or congestion level

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Software Systems (AREA)
  • Artificial Intelligence (AREA)
  • Mathematical Physics (AREA)
  • General Physics & Mathematics (AREA)
  • Computational Linguistics (AREA)
  • Computing Systems (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • General Health & Medical Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Environmental & Geological Engineering (AREA)
  • Databases & Information Systems (AREA)
  • Medical Informatics (AREA)
  • Algebra (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Optimization (AREA)
  • Probability & Statistics with Applications (AREA)
  • Pure & Applied Mathematics (AREA)
  • Data Exchanges In Wide-Area Networks (AREA)

Abstract

The disclosure provides a method, a device, equipment and a storage medium for predicting network traffic and training a model, which relate to the technical field of artificial intelligence, in particular to the technical fields of deep learning, internet and the like. The network traffic prediction method comprises the following steps: acquiring historical flow data of a plurality of network nodes; processing the historical flow data based on the association relation between every two nodes in the plurality of network nodes to obtain fusion characteristics; future traffic data of the plurality of network nodes is determined based on the fusion characteristics. The method and the device can improve accuracy of driving network flow prediction.

Description

Network traffic prediction and model training method, device, equipment and storage medium
Technical Field
The disclosure relates to the technical field of artificial intelligence, in particular to the technical fields of deep learning, internet and the like, and particularly relates to a method, a device, equipment and a storage medium for predicting network traffic and training a model.
Background
With the development of internet technology, the network scale is continuously enlarged, the network traffic is greatly increased, and the phenomenon of congestion or overload is frequently caused in the network. The network traffic can be accurately predicted, the network can be effectively managed, and the comprehensive performance of the network is improved.
Disclosure of Invention
The present disclosure provides a method, apparatus, device and storage medium for predicting network traffic and training a model.
According to an aspect of the present disclosure, there is provided a network traffic prediction method, including: acquiring historical flow data of a plurality of network nodes; processing the historical flow data based on the association relation between every two nodes in the plurality of network nodes to obtain fusion characteristics; future traffic data of the plurality of network nodes is determined based on the fusion characteristics.
According to another aspect of the present disclosure, there is provided a training method of a network traffic prediction model, the model including: a graph-rolling network, the method comprising: obtaining sample data for a plurality of network nodes, the sample data comprising: historical traffic data and real future traffic data; adopting the graph convolution network to process the historical flow data based on the association relationship between every two nodes in the plurality of network nodes so as to obtain fusion characteristics; based on the fusion characteristics, obtaining predicted future flow data; constructing a loss function based on the real future flow data and the predicted future flow data; based on the loss function, adjusting model parameters until a preset condition is met, wherein the model parameters comprise: the graph is rolled up with model parameters of the network.
According to another aspect of the present disclosure, there is provided a network traffic prediction apparatus, including: the acquisition module is used for acquiring historical flow data of a plurality of network nodes; the fusion module is used for processing the historical flow data based on the association relation between every two nodes in the plurality of network nodes so as to obtain fusion characteristics; and the prediction module is used for determining future flow data of the plurality of network nodes based on the fusion characteristics.
According to another aspect of the present disclosure, there is provided a training apparatus of a network traffic prediction model, the model including: a graph-rolling network, the apparatus comprising: an acquisition module, configured to acquire sample data of a plurality of network nodes, where the sample data includes: historical traffic data and real future traffic data; the fusion module is used for processing the historical flow data based on the association relation between every two nodes in the plurality of network nodes by adopting the graph rolling network so as to obtain fusion characteristics; the prediction module is used for obtaining predicted future flow data based on the fusion characteristics; the construction module is used for constructing a loss function based on the real future flow data and the predicted future flow data; the adjusting module is configured to adjust model parameters based on the loss function until a preset condition is satisfied, where the model parameters include: the graph is rolled up with model parameters of the network.
According to another aspect of the present disclosure, there is provided an electronic device including: at least one processor; and a memory communicatively coupled to the at least one processor; wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method of any one of the above aspects.
According to another aspect of the present disclosure, there is provided a non-transitory computer readable storage medium storing computer instructions for causing the computer to perform the method according to any one of the above aspects.
According to another aspect of the present disclosure, there is provided a computer program product comprising a computer program which, when executed by a processor, implements a method according to any of the above aspects.
According to the technical scheme, the accuracy of network traffic prediction can be improved.
It should be understood that the description in this section is not intended to identify key or critical features of the embodiments of the disclosure, nor is it intended to be used to limit the scope of the disclosure. Other features of the present disclosure will become apparent from the following specification.
Drawings
The drawings are for a better understanding of the present solution and are not to be construed as limiting the present disclosure. Wherein:
FIG. 1 is a schematic diagram according to a first embodiment of the present disclosure;
FIG. 2 is a schematic illustration of a micro-service scenario provided in accordance with an embodiment of the present disclosure;
FIG. 3 is a schematic diagram of an overall architecture of network traffic prediction provided in accordance with an embodiment of the present disclosure;
FIG. 4 is a schematic diagram according to a second embodiment of the present disclosure;
FIG. 5 is a schematic diagram according to a third embodiment of the present disclosure;
FIG. 6 is a schematic diagram according to a fourth embodiment of the present disclosure;
FIG. 7 is a schematic diagram according to a fifth embodiment of the present disclosure;
fig. 8 is a schematic diagram of an electronic device used to implement a network traffic prediction method or a training method of a network traffic prediction model of an embodiment of the present disclosure.
Detailed Description
Exemplary embodiments of the present disclosure are described below in conjunction with the accompanying drawings, which include various details of the embodiments of the present disclosure to facilitate understanding, and should be considered as merely exemplary. Accordingly, one of ordinary skill in the art will recognize that various changes and modifications of the embodiments described herein can be made without departing from the scope and spirit of the present disclosure. Also, descriptions of well-known functions and constructions are omitted in the following description for clarity and conciseness.
In the related art, some schemes of network traffic prediction exist, but the accuracy needs to be improved.
Fig. 1 is a schematic diagram of a first embodiment of the present disclosure, where the present embodiment provides a network traffic prediction method, and the method includes:
101. historical traffic data for a plurality of network nodes is obtained.
102. And processing the historical flow data based on the association relation between every two nodes in the plurality of network nodes to obtain fusion characteristics.
103. Future traffic data of the plurality of network nodes is determined based on the fusion characteristics.
Wherein the plurality of network nodes are nodes having an association relationship, for example, nodes having a call relationship.
Assuming that the number of the plurality of network nodes is N (positive integer), every two nodes can be represented by an ith node and a jth node, i, j= [1, N]The association relation can be A ij And (3) representing.
Wherein A is ij Is located at [0,1 ]]The value in between can be determined in particular by a training process.
Its initial value may be 0 or 1, e.g., there is a call relationship between the ith node and the jth node, then A ij =1, otherwise, if there is no call relationship between the two, a ij ＝0。
The fusion feature is obtained based on historical traffic data, and integrates information of a plurality of network nodes.
After the fusion features are obtained, network traffic may be predicted based on the fusion features.
In the related art, N models can be used to predict network traffic of each node for N network nodes, and because the network traffic is processed for each node, and the association relationship between the nodes is not considered, however, in some scenarios, when there is an association relationship between a plurality of nodes, the network traffic is affected, and only information of a single node is predicted independently, so that there is a problem of insufficient prediction accuracy.
In this embodiment, the fusion feature is obtained based on the association relationship, and the future traffic data is determined based on the fusion feature, so that the association relationship between nodes can be considered in network traffic prediction, traffic prediction can be performed based on the fusion information, and the accuracy of network traffic prediction can be improved.
In order to better understand the embodiments of the present disclosure, application scenarios to which the embodiments of the present disclosure are applicable are described below.
The embodiment can be applied to a micro-service scene. A microservice (or microservice architecture) is a service-oriented architecture, where a single application consists of many loosely coupled and independently deployable smaller services, which may be referred to as microservices. Each micro-service is built around a specific business and can be deployed independently into a production environment, class production environment.
As shown in fig. 2, a client may be deployed on a user terminal 201, and a micro service may be deployed on a server 202, where the user terminal includes, for example: personal computers (personalcomputers), notebook computers, mobile devices (e.g., cell phones), and the like. The server may be a local server or a cloud server. The user terminal and the server may communicate over a wired network and/or a wireless network.
The micro services are usually multiple, and are represented by micro services 1 to micro service n in fig. 2, each micro service may be deployed on a corresponding network node, and are represented by nodes 1 to node n in fig. 2, where the nodes may be servers or processors, etc.
In a micro-service scenario, there may be a call relationship between different network nodes, for example, the micro-service includes: the order service is used for generating an order, the user service is used for providing user information, the commodity service is used for providing commodity information, the order requires the user information and the commodity information, the network node to which the commodity service belongs is assumed to be a node A, the network node to which the commodity service belongs is assumed to be a node B, the network node to which the order service belongs is a node C, and the node C needs to call the node A and the node B, namely, the node C has a call relationship (or is called an association relationship) with the node A, and the node C has a call relationship with the node B.
In the related art, N models can be used for predicting network traffic for N network nodes, but in a micro service scenario, because of the association relationship between different network nodes, a mode of predicting network traffic by using one model has a problem of insufficient accuracy.
In order to improve the accuracy of network traffic prediction, the embodiment of the disclosure will consider association information between different network nodes.
As shown in fig. 3, the overall architecture of network traffic prediction mainly includes: the fusion feature extraction network G and the time feature extraction network M are used for extracting fusion features based on association relations among network nodes, wherein the fusion features refer to features that one network node fuses other network nodes. The fused feature extraction network may be a graph roll-up network (GraphConvolutional Network), which in fig. 3 includes two graph roll-up layers. The temporal feature extraction network is used to obtain temporal features based on the fused features, and the temporal feature extraction network may be a recurrent neural network (RecurrentNeuralNetwork, RNN), specifically a long short-term memory (LSTM) network in fig. 3.
In addition, as shown in fig. 3, the overall architecture may further include two fully-connected layers, which are respectively represented by a first fully-connected layer and a second fully-connected layer for distinguishing, where the first fully-connected layer is used for normalizing historical traffic data of N nodes, and the second fully-connected layer is used for processing time features to obtain predicted future traffic data of N nodes.
The network traffic prediction model specifically includes the following: a first fully connected layer, a graph roll-up network, an LSTM network, and a second fully connected layer. The model can be trained in the training phase, and can be used for network traffic prediction in the reasoning phase.
In combination with the application scenario, the disclosure further provides a network traffic prediction method.
Fig. 4 is a schematic diagram of a second embodiment of the present disclosure, where the present embodiment provides a network traffic prediction method, and the method includes:
401. historical traffic data for a plurality of network nodes is obtained.
Where, assuming that the number of network nodes is represented by N (positive integer), historical traffic data for N network nodes may be obtained.
For each network node, the historical traffic data comprises traffic data of a preset number of continuous time points, and the preset number is T (positive integer), and the historical traffic data comprises traffic data of the current moment (T moment) and the previous moment (T-1). Wherein, by sampling with a preset period (such as 1 minute), flow data at each moment can be obtained.
Suppose that X is used for historical flow data 0 X is represented by 0 ∈R N×T . Namely, input data X 0 Each row corresponds to a network node, and each row data is historical flow data of the corresponding node at T moments.
402. And carrying out normalization processing on the historical flow data by adopting a first full-connection layer so as to obtain normalized historical flow data.
The historical traffic data of different network nodes may be greatly different, for example, 10 times or 1 ten thousand times, and the historical traffic data of each network node can be normalized to be within the range of [0,1] through normalization processing.
In connection with fig. 3, the normalization process may be performed using the first full connection layer.
Specifically, the initial historical flow data X 0 Is input into a first full connection layer, and the first full connection layer inputs data X 0 After processing, the data is output as normalized historical flow data X.
Wherein the output and input of the first fully connected layer are of the same dimension, i.e. X ε R N×T 。
The first fully connected layer is obtained by training.
In this embodiment, by performing normalization processing on the historical traffic data of the plurality of network nodes, the historical traffic data of the plurality of network nodes can be normalized to a set interval, so that excessive numerical value differences are avoided, and further, the accuracy of network traffic prediction can be improved.
403. Processing the normalized historical flow data by adopting a pre-trained graph convolution network to obtain fusion characteristics; wherein the parameters of the graph rolling network include: and the adjacency matrix is used for representing the association relation between every two nodes in the plurality of network nodes.
Wherein, in connection with fig. 3, the graph convolution network comprises two graph convolution layers, and parameters related to each graph convolution layer comprise: an adjacency matrix A and a weight parameter matrix W.
Specifically, the normalized historical flow data may be input into a graph convolution network, and the graph convolution network processes the input data and outputs the processed input data as a fusion feature.
The calculation formula for a single graph convolutional layer can be expressed as:
H l+1 ＝σ(AH l W l )
wherein H is l+1 Is the output of the first layer; h l Is an input of the first layer, the initial value (i.e. the input of the first layer) is H 1 ＝X；W l Is the weight parameter matrix of the first layer; sigma () represents an activation function, which may specifically be a ReLU activation function; a is an adjacency matrix.
Wherein A is E R N×N I.e. a matrix of dimension N x N, for element a therein ij Representing the association relationship between the ith network node and the jth network node, i, j E [1, N]。
The graph-rolling network is obtained by training, i.e. a and W as described above can be obtained by training.
The dimension of the output data of the graph rolling network is the same as the dimension of the input data, i.e. a matrix with dimension N x T.
In this embodiment, the graph convolution network is used to process the historical traffic data, and because the parameters of the graph convolution network include the adjacency matrix, the adjacency matrix is used to represent the association relationship between every two nodes, so that the fusion characteristic of fusing the information of multiple network nodes can be obtained, and the prediction accuracy is further improved.
404. And processing the fusion characteristics by adopting a pre-trained time characteristic extraction network to obtain time characteristics.
Wherein, in connection with fig. 3, the temporal feature extraction network may be an LSTM network.
Specifically, the fusion feature may be input into an LSTM network, and the LSTM network processes the input data and outputs the processed data as a temporal feature.
The LSTM network may include two layers, where the number of output nodes in each layer may be denoted by K (positive integer), and K may be set manually. Thus, the dimension of the output data (time characteristics) of the LSTM network is an nxk matrix.
In this embodiment, the time feature is obtained after the fusion feature is extracted by using the time feature extraction network, so that the time feature has richer information, and when future flow data is obtained based on the time feature, the richer information can be used for predicting the network flow, thereby improving the prediction accuracy.
405. And processing the time characteristics by adopting a second full connection layer to obtain future flow data of the plurality of network nodes.
For each network node, the future traffic data may be specifically traffic data of the next time, and if the current time is the T-th time, the future time is the (t+1) -th time, and the traffic data of the (t+1) -th time is predicted.
Wherein future traffic data for N network nodes may be obtained simultaneously.
Thus, assuming future flow data is represented by Y, Y ε R N×1 . That is, the output data Y is an N-dimensional vector, each dimension corresponds to a network node, and the data in each dimension is historical traffic data at time t+1 of the corresponding node.
The above-described prediction process may be continuous, i.e., the flow data at the (t+1) th time is predicted from the flow data at the 1 st to T times, the flow data at the (t+2) th time is predicted from the flow data at the 2 nd to (t+1) th times, and so on.
In this embodiment, the second full connection layer may convert the time feature into the corresponding future traffic data more succinctly, so as to improve the efficiency of network traffic prediction.
Furthermore, the network traffic prediction model of the embodiment is end-to-end, the input is historical traffic data, the output is a prediction result, namely future traffic data, no special structure is required to be additionally introduced, the operation can be simplified, and the efficiency is improved. The network flow prediction model is uniformly used by a plurality of network nodes, can synchronously predict the network flows of N network nodes, does not need to use a separate model for each network node, and can simplify the system structure and improve the prediction accuracy. By adopting the graph convolution network, the information of a plurality of network nodes can be fused, and prediction is performed based on the fused information, so that the prediction accuracy can be improved.
The above embodiment relates to a network traffic prediction model, and a model training process is described below.
Fig. 5 is a schematic diagram of a third embodiment of the present disclosure, where the present embodiment provides a training method of a network traffic prediction model, the model includes: a graph-rolling network, the method comprising:
501. obtaining sample data for a plurality of network nodes, the sample data comprising: historical traffic data, and real future traffic data.
502. And processing the historical flow data based on the association relation between every two nodes in the plurality of network nodes by adopting the graph rolling network to obtain fusion characteristics.
503. And obtaining predicted future flow data based on the fusion characteristics.
504. A loss function is constructed based on the real future flow data and the predicted future flow data.
505. Based on the loss function, adjusting model parameters until a preset condition is met, wherein the model parameters comprise: the graph is rolled up with model parameters of the network.
In this embodiment, the fusion feature is obtained by adopting the graph convolution network, and because the fusion feature is obtained based on the association relationship between every two nodes in the plurality of network nodes, the plurality of network node information is fused, so that the model constructed based on the fusion feature fuses the plurality of node information, the accuracy of the model can be improved, and the prediction accuracy can be improved by performing network traffic prediction based on the model.
In some embodiments, the model comprises: a feature extraction network, the model parameters further comprising: the feature extracting network model parameters, based on the fusion feature, obtain predicted future flow data, including: processing the fusion features by adopting the feature extraction network to obtain time features; the predicted future flow characteristic is obtained based on the temporal characteristic.
In this embodiment, the time feature is obtained after the fusion feature is extracted by using the time feature extraction network, so that the time feature has richer information, and when future flow data is obtained based on the time feature, the richer information can be used for predicting the network flow, thereby improving the prediction accuracy.
In some embodiments, the model further comprises: the first full connection layer, the model parameters further include: model parameters of the first fully connected layer, the method further comprising: carrying out normalization processing on the historical flow data by adopting the first full connection layer to obtain normalized historical flow data; correspondingly, the processing the historical traffic data by adopting the graph rolling network to obtain the fusion characteristic comprises the following steps: and processing the normalized historical flow data by adopting the graph convolution network to obtain the fusion characteristic.
In this embodiment, by performing normalization processing on the historical traffic data of the plurality of network nodes, the historical traffic data of the plurality of network nodes can be normalized to a set interval, so that excessive numerical value differences are avoided, and the accuracy of the network traffic prediction model can be improved.
In some embodiments, the model further comprises: a second fully connected layer, the model parameters further comprising: the model parameters of the second fully connected layer, the obtaining the predicted future flow characteristic based on the temporal characteristic, includes: and processing the time characteristic by adopting the second full connection layer to obtain the predicted future flow data.
In this embodiment, the second full connection layer may convert the time feature into the corresponding future traffic data more succinctly, so as to improve the efficiency of network traffic prediction.
In some embodiments, the obtaining sample data for a plurality of network nodes includes:
acquiring time point flow data of a preset number of a plurality of network nodes;
dividing the time point flow data of the preset number into the historical flow data and the real future flow data; wherein the real future flow data is last time point flow data, and the historical flow data comprises: and other time point flow data except the last time point flow data.
For N network nodes, each network node may obtain flow data of (t+1) continuous time points, take the flow data of the (t+1) th time point as real future flow data, and take the rest flow data, i.e. the flow data of the 1 st to T time points as historical flow data. Thus, the dimension of the historical traffic data is n×t, and the dimension of the real future traffic data is N.
In this embodiment, by performing packet processing on the preset number of time point traffic data, the historical traffic data and the real future traffic data in the sample data can be simply, conveniently and quickly obtained.
After the model is adopted to obtain the predicted future flow data, a loss function can be constructed based on the predicted future flow data and the real future flow data, and the loss function can be specifically a Mean square error (Mean SquareError, MSE) function.
After the loss function is obtained, the model parameters can be adjusted based on the loss function until the preset condition is met, and a final model is obtained. For example, a back propagation (BackPropagation, BP) algorithm may be employed to adjust the model parameters.
The adjustment parameters may be: the adjusted parameter=parameter before adjustment-learning rate-gradient value, wherein the gradient value may be obtained after the loss function derives from the model parameter, and the learning rate is a preset super parameter. The initial values of the respective parameters may be set, for example, randomization processing may be performed for the respective parameters, and the initial values of the respective parameters may be obtained at random.
For the first full-connection layer, the time feature extraction network and the second full-connection layer, initial values of model parameters of weight parameter matrixes W and W can be randomly generated.
For the graph convolutional network, the model parameters of the graph convolutional network comprise a weight parameter matrix W and an adjacent matrix A, initial values of the weight parameter matrix W can be randomly generated, the initial values of the A can be determined according to scheduling relations among network nodes, for example, for Aij, if a calling relation exists between the node i and the node j, the Aij=1, and otherwise, the Aij=0. And then, the A can be subjected to learning adjustment in the training process, and the element Aij in the final A is a value between 0 and 1.
The preset condition is, for example, that the number of iterations reaches a preset value or a preset convergence condition is satisfied, for example, that the absolute value of the difference between two adjacent loss functions is smaller than the preset value. The model meeting the preset conditions is taken as a final network flow prediction model, and comprises the following steps: the system comprises a first full connection layer, a graph rolling network, a time feature extraction network and a second full connection layer. The final network traffic prediction model may be used for the inference phase of network traffic prediction.
In this embodiment, an end-to-end model can be obtained through the training process, and the model can be used for predicting network traffic in an inference stage, so that compared with a mode of additionally designing or a plurality of models, the complexity can be reduced, the inference speed can be improved, in addition, the information of a plurality of network nodes is fused in the training process, the model accuracy can be improved, and further the inference accuracy is improved.
Fig. 6 is a schematic diagram of a third embodiment of the present disclosure, where a network traffic prediction apparatus 600 includes: an acquisition module 601, a fusion module 602 and a prediction module 603.
The acquiring module 601 is configured to acquire historical traffic data of a plurality of network nodes; the fusion module 602 is configured to process the historical traffic data based on association relationships between two nodes in the plurality of network nodes, so as to obtain fusion features; the prediction module 603 is configured to determine future traffic data of the plurality of network nodes based on the fusion characteristics.
In this embodiment, by obtaining the fusion feature based on the association relationship, and determining future traffic data based on the fusion feature, the association relationship between nodes can be considered in network traffic prediction, and traffic prediction can be performed based on the fusion information, so that accuracy of network traffic prediction can be improved.
In some embodiments, the fusion module 602 is further configured to: processing the historical flow by adopting a pre-trained graph rolling network to obtain the fusion characteristic; wherein the parameters of the graph rolling network include: and the adjacency matrix is used for representing the association relation between every two nodes.
In this embodiment, the graph convolution network is used to process the historical traffic data, and because the parameters of the graph convolution network include the adjacency matrix, the adjacency matrix is used to represent the association relationship between every two nodes, so that the fusion characteristic of fusing the information of multiple network nodes can be obtained, and the prediction accuracy is further improved.
In some embodiments, the prediction module 603 is further configured to: processing the fusion features by adopting a pre-trained time feature extraction network to obtain time features; based on the temporal characteristics, the future flow data is obtained.
In this embodiment, the time feature is obtained after the fusion feature is extracted by using the time feature extraction network, so that the time feature has richer information, and when future flow data is obtained based on the time feature, the richer information can be used for predicting the network flow, thereby improving the prediction accuracy.
In some embodiments, the apparatus 600 further comprises: and a normalization module. The normalization module is used for performing normalization processing on the historical flow data by adopting a first full-connection layer trained in advance so as to obtain normalized historical flow data; accordingly, the fusion module 602 is further configured to: and processing the normalized historical flow data based on the association relation between every two nodes in the plurality of network nodes to obtain fusion characteristics.
In this embodiment, by performing normalization processing on the historical traffic data of the plurality of network nodes, the historical traffic data of the plurality of network nodes can be normalized to a set interval, so that excessive numerical value differences are avoided, and further, the accuracy of network traffic prediction can be improved.
In some embodiments, the prediction module 603 is further configured to: and processing the time characteristic by adopting a pre-trained second full connection layer to obtain the future flow data.
In this embodiment, the second full connection layer may convert the time feature into the corresponding future traffic data more succinctly, so as to improve the efficiency of network traffic prediction.
Fig. 7 is a schematic diagram of a fourth embodiment of the present disclosure, where the present embodiment provides a training apparatus for a network traffic prediction model, the model includes: the graph packing network the apparatus 700 includes: an acquisition module 701, a fusion module 702, a prediction module 703, a construction module 704 and an adjustment module 705.
The obtaining module 701 is configured to obtain sample data of a plurality of network nodes, where the sample data includes: historical traffic data and real future traffic data; the fusion module 702 is configured to process the historical traffic data based on the association relationship between every two nodes in the plurality of network nodes by using the graph rolling network, so as to obtain a fusion feature; the prediction module 703 is configured to obtain predicted future flow data based on the fusion feature; a construction module 704 for constructing a loss function based on the real future flow data and the predicted future flow data; the adjustment module 705 is configured to adjust model parameters based on the loss function until a preset condition is met, where the model parameters include: the graph is rolled up with model parameters of the network.
In this embodiment, the fusion feature is obtained by adopting the graph convolution network, and because the fusion feature is obtained based on the association relationship between every two nodes in the plurality of network nodes, the plurality of network node information is fused, so that the model constructed based on the fusion feature fuses the plurality of node information, the accuracy of the model can be improved, and the prediction accuracy can be improved by performing network traffic prediction based on the model.
In some embodiments, the model comprises: a feature extraction network, the model parameters further comprising: the prediction module 703 is further configured to: processing the fusion features by adopting the feature extraction network to obtain time features; the predicted future flow characteristic is obtained based on the temporal characteristic.
In this embodiment, the time feature is obtained after the fusion feature is extracted by using the time feature extraction network, so that the time feature has richer information, and when future flow data is obtained based on the time feature, the richer information can be used for predicting the network flow, thereby improving the prediction accuracy.
In some embodiments, the model further comprises: the first full connection layer, the model parameters further include: model parameters of the first fully connected layer, the apparatus further comprising: the normalization module is used for carrying out normalization processing on the historical flow data by adopting the first full-connection layer so as to obtain normalized historical flow data; accordingly, the fusion module 703 is further configured to: and processing the normalized historical flow data by adopting the graph convolution network to obtain fusion characteristics.
In this embodiment, by performing normalization processing on the historical traffic data of the plurality of network nodes, the historical traffic data of the plurality of network nodes can be normalized to a set interval, so that excessive numerical value differences are avoided, and the accuracy of the network traffic prediction model can be improved.
In some embodiments, the model further comprises: a second fully connected layer, the model parameters further comprising: the prediction module 703 is further configured to: and processing the time characteristic by adopting the second full connection layer to obtain the predicted future flow data.
In this embodiment, the second full connection layer may convert the time feature into the corresponding future traffic data more succinctly, so as to improve the efficiency of network traffic prediction.
The obtaining module 701 is further configured to: acquiring time point flow data of a preset number of a plurality of network nodes; dividing the time point flow data of the preset number into the historical flow data and the real future flow data; wherein the real future flow data is last time point flow data, and the historical flow data comprises: and other time point flow data except the last time point flow data.
In this embodiment, by performing packet processing on the preset number of time point traffic data, the historical traffic data and the real future traffic data in the sample data can be simply, conveniently and quickly obtained.
It is to be understood that in the embodiments of the disclosure, the same or similar content in different embodiments may be referred to each other.
It can be understood that "first", "second", etc. in the embodiments of the present disclosure are only used for distinguishing, and do not indicate the importance level, the time sequence, etc.
In the technical scheme of the disclosure, the related processes of collecting, storing, using, processing, transmitting, providing, disclosing and the like of the personal information of the user accord with the regulations of related laws and regulations, and the public order colloquial is not violated.
According to embodiments of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium and a computer program product.
Fig. 8 illustrates a schematic block diagram of an example electronic device 800 that may be used to implement embodiments of the present disclosure. Electronic device 800 is intended to represent various forms of digital computers, such as laptops, desktops, workstations, servers, blade servers, mainframes, and other appropriate computers. Electronic device 800 may also represent various forms of mobile devices, such as personal digital assistants, cellular telephones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions, are meant to be exemplary only, and are not meant to limit implementations of the disclosure described and/or claimed herein.
As shown in fig. 8, the electronic device 800 includes a computing unit 801 that can perform various appropriate actions and processes according to a computer program stored in a Read Only Memory (ROM) 802 or a computer program loaded from a storage unit 808 into a Random Access Memory (RAM) 803. In the RAM803, various programs and data required for the operation of the electronic device 800 can also be stored. The computing unit 801, the ROM802, and the RAM803 are connected to each other by a bus 804. An input/output (I/O) interface 805 is also connected to the bus 804.
Various components in electronic device 800 are connected to I/O interface 805, including: an input unit 806 such as a keyboard, mouse, etc.; an output unit 807 such as various types of displays, speakers, and the like; a storage unit 808, such as a magnetic disk, optical disk, etc.; and a communication unit 809, such as a network card, modem, wireless communication transceiver, or the like. The communication unit 809 allows the electronic device 800 to exchange information/data with other devices through a computer network such as the internet and/or various telecommunication networks.
The computing unit 801 may be a variety of general and/or special purpose processing components having processing and computing capabilities. Some examples of computing unit 801 include, but are not limited to, a Central Processing Unit (CPU), a Graphics Processing Unit (GPU), various specialized Artificial Intelligence (AI) computing chips, various computing units running machine learning model algorithms, a Digital Signal Processor (DSP), and any suitable processor, controller, microcontroller, etc. The calculation unit 801 performs the respective methods and processes described above, for example, a network traffic prediction method or a training method of a network traffic prediction model. For example, in some embodiments, the network traffic prediction method or the training method of the network traffic prediction model may be implemented as a computer software program tangibly embodied on a machine-readable medium, such as the storage unit 808. In some embodiments, part or all of the computer program may be loaded and/or installed onto the electronic device 800 via the ROM802 and/or the communication unit 809. When the computer program is loaded into RAM803 and executed by computing unit 801, one or more steps of the above-described network traffic prediction method or training method of the network traffic prediction model may be performed. Alternatively, in other embodiments, the computing unit 801 may be configured to perform the network traffic prediction method or the training method of the network traffic prediction model in any other suitable way (e.g., by means of firmware).
Various implementations of the systems and techniques described here above may be implemented in digital electronic circuitry, integrated circuitry, field Programmable Gate Arrays (FPGAs), application Specific Integrated Circuits (ASICs), application Specific Standard Products (ASSPs), systems-on-chips (SOCs), complex Programmable Logic Devices (CPLDs), computer hardware, firmware, software, and/or combinations thereof. These various embodiments may include: implemented in one or more computer programs, the one or more computer programs may be executed and/or interpreted on a programmable system including at least one programmable processor, which may be a special purpose or general-purpose programmable processor, that may receive data and instructions from, and transmit data and instructions to, a storage system, at least one input device, and at least one output device.
Program code for carrying out methods of the present disclosure may be written in any combination of one or more programming languages. These program code may be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable load balancing apparatus, such that the program code, when executed by the processor or controller, causes the functions/operations specified in the flowchart and/or block diagram to be implemented. The program code may execute entirely on the machine, partly on the machine, as a stand-alone software package, partly on the machine and partly on a remote machine or entirely on the remote machine or server.
In the context of this disclosure, a machine-readable medium may be a tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. The machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.
To provide for interaction with a user, the systems and techniques described here can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to a user; and a keyboard and pointing device (e.g., a mouse or trackball) by which a user can provide input to the computer. Other kinds of devices may also be used to provide for interaction with a user; for example, feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form, including acoustic input, speech input, or tactile input.
The systems and techniques described here can be implemented in a computing system that includes a background component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front-end component (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such background, middleware, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: local Area Networks (LANs), wide Area Networks (WANs), and the internet.
The computer system may include a client and a server. The client and server are typically remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. The server can be a cloud server, also called a cloud computing server or a cloud host, and is a host product in a cloud computing service system, so that the defects of high management difficulty and weak service expansibility in the traditional physical hosts and VPS service ("VirtualPrivate Server" or simply "VPS") are overcome. The server may also be a server of a distributed system or a server that incorporates a blockchain.
It should be appreciated that various forms of the flows shown above may be used to reorder, add, or delete steps. For example, the steps recited in the present disclosure may be performed in parallel or sequentially or in a different order, provided that the desired results of the technical solutions of the present disclosure are achieved, and are not limited herein.
The above detailed description should not be taken as limiting the scope of the present disclosure. It will be apparent to those skilled in the art that various modifications, combinations, sub-combinations and alternatives are possible, depending on design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present disclosure are intended to be included within the scope of the present disclosure.

Claims (23)

1. A network traffic prediction method, comprising:
acquiring historical flow data of a plurality of network nodes;
processing the historical flow data based on the association relation between every two nodes in the plurality of network nodes to obtain fusion characteristics;
future traffic data of the plurality of network nodes is determined based on the fusion characteristics.
2. The method of claim 1, wherein the processing the historical traffic data based on the association between two nodes in the plurality of network nodes to obtain a fusion feature comprises:
Processing the historical flow by adopting a pre-trained graph rolling network to obtain the fusion characteristic; wherein the parameters of the graph rolling network include: and the adjacency matrix is used for representing the association relation between every two nodes.
3. The method of claim 1, wherein the determining future traffic data for the plurality of network nodes based on the fusion characteristics comprises:
processing the fusion features by adopting a pre-trained time feature extraction network to obtain time features;
based on the temporal characteristics, the future flow data is obtained.
4. A method according to any one of claims 1-3, further comprising:
carrying out normalization processing on the historical flow data by adopting a first full-connection layer trained in advance so as to obtain normalized historical flow data;
correspondingly, the processing the historical traffic data based on the association relationship between every two nodes in the plurality of network nodes to obtain the fusion feature comprises the following steps:
and processing the normalized historical flow data based on the association relation between every two nodes in the plurality of network nodes to obtain fusion characteristics.
5. A method according to claim 3, wherein said obtaining said future flow data based on said temporal characteristics comprises:
and processing the time characteristic by adopting a pre-trained second full connection layer to obtain the future flow data.
6. A method of training a network traffic prediction model, the model comprising: a graph-rolling network, the method comprising:
obtaining sample data for a plurality of network nodes, the sample data comprising: historical traffic data and real future traffic data;
adopting the graph convolution network to process the historical flow data based on the association relationship between every two nodes in the plurality of network nodes so as to obtain fusion characteristics;
based on the fusion characteristics, obtaining predicted future flow data;
constructing a loss function based on the real future flow data and the predicted future flow data;
based on the loss function, adjusting model parameters until a preset condition is met, wherein the model parameters comprise: the graph is rolled up with model parameters of the network.
7. The method of claim 6, wherein the model comprises: a feature extraction network, the model parameters further comprising: the feature extracting network model parameters, based on the fusion feature, obtain predicted future flow data, including:
Processing the fusion features by adopting the feature extraction network to obtain time features;
the predicted future flow characteristic is obtained based on the temporal characteristic.
8. The method of claim 6, wherein the model further comprises: the first full connection layer, the model parameters further include: model parameters of the first fully connected layer, the method further comprising:
carrying out normalization processing on the historical flow data by adopting the first full connection layer to obtain normalized historical flow data;
correspondingly, the processing the historical traffic data by adopting the graph rolling network to obtain the fusion characteristic comprises the following steps: and processing the normalized historical flow data by adopting the graph convolution network to obtain the fusion characteristic.
9. The method of claim 7, wherein the model further comprises: a second fully connected layer, the model parameters further comprising: the model parameters of the second fully connected layer, the obtaining the predicted future flow characteristic based on the temporal characteristic, includes:
and processing the time characteristic by adopting the second full connection layer to obtain the predicted future flow data.
10. The method according to any of claims 6-9, wherein the obtaining sample data for a plurality of network nodes comprises:
acquiring time point flow data of a preset number of a plurality of network nodes;
dividing the time point flow data of the preset number into the historical flow data and the real future flow data; wherein the real future flow data is last time point flow data, and the historical flow data comprises: and other time point flow data except the last time point flow data.
11. A network traffic prediction device, comprising:
the acquisition module is used for acquiring historical flow data of a plurality of network nodes;
the fusion module is used for processing the historical flow data based on the association relation between every two nodes in the plurality of network nodes so as to obtain fusion characteristics;
and the prediction module is used for determining future flow data of the plurality of network nodes based on the fusion characteristics.
12. The apparatus of claim 11, wherein the fusion module is further to:
processing the historical flow by adopting a pre-trained graph rolling network to obtain the fusion characteristic; wherein the parameters of the graph rolling network include: and the adjacency matrix is used for representing the association relation between every two nodes.
13. The apparatus of claim 11, wherein the prediction module is further to:
processing the fusion features by adopting a pre-trained time feature extraction network to obtain time features;
based on the temporal characteristics, the future flow data is obtained.
14. The apparatus of any of claims 11-13, further comprising:
the normalization module is used for performing normalization processing on the historical flow data by adopting a first full-connection layer trained in advance so as to obtain normalized historical flow data;
accordingly, the fusion module is further configured to: and processing the normalized historical flow data based on the association relation between every two nodes in the plurality of network nodes to obtain fusion characteristics.
15. The apparatus of claim 13, wherein the prediction module is further to:
and processing the time characteristic by adopting a pre-trained second full connection layer to obtain the future flow data.
16. A training device for a network traffic prediction model, the model comprising: a graph-rolling network, the apparatus comprising:
an acquisition module, configured to acquire sample data of a plurality of network nodes, where the sample data includes: historical traffic data and real future traffic data;
The fusion module is used for processing the historical flow data based on the association relation between every two nodes in the plurality of network nodes by adopting the graph rolling network so as to obtain fusion characteristics;
the prediction module is used for obtaining predicted future flow data based on the fusion characteristics;
the construction module is used for constructing a loss function based on the real future flow data and the predicted future flow data;
the adjusting module is configured to adjust model parameters based on the loss function until a preset condition is satisfied, where the model parameters include: the graph is rolled up with model parameters of the network.
17. The apparatus of claim 16, wherein the model comprises: a feature extraction network, the model parameters further comprising: the model parameters of the feature extraction network, the prediction module further to:
processing the fusion features by adopting the feature extraction network to obtain time features;
the predicted future flow characteristic is obtained based on the temporal characteristic.
18. The apparatus of claim 16, wherein the model further comprises: the first full connection layer, the model parameters further include: model parameters of the first fully connected layer, the apparatus further comprising:
The normalization module is used for carrying out normalization processing on the historical flow data by adopting the first full-connection layer so as to obtain normalized historical flow data;
accordingly, the fusion module is further configured to: and processing the normalized historical flow data by adopting the graph convolution network to obtain fusion characteristics.
19. The apparatus of claim 17, wherein the model further comprises: a second fully connected layer, the model parameters further comprising: the model parameters of the second fully connected layer, the prediction module further to:
and processing the time characteristic by adopting the second full connection layer to obtain the predicted future flow data.
20. The apparatus of any of claims 16-19, wherein the acquisition module is further to:
acquiring time point flow data of a preset number of a plurality of network nodes;
dividing the time point flow data of the preset number into the historical flow data and the real future flow data; wherein the real future flow data is last time point flow data, and the historical flow data comprises: and other time point flow data except the last time point flow data.
21. An electronic device, comprising:
at least one processor; and
a memory communicatively coupled to the at least one processor; wherein,,
the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-10.
22. A non-transitory computer readable storage medium storing computer instructions for causing the computer to perform the method of any one of claims 1-10.
23. A computer program product comprising a computer program which, when executed by a processor, implements the method according to any of claims 1-10.
CN202310228544.2A 2023-03-03 2023-03-03 Network traffic prediction and model training method, device, equipment and storage medium Pending CN116436828A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310228544.2A CN116436828A (en) 2023-03-03 2023-03-03 Network traffic prediction and model training method, device, equipment and storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310228544.2A CN116436828A (en) 2023-03-03 2023-03-03 Network traffic prediction and model training method, device, equipment and storage medium

Publications (1)

Publication Number Publication Date
CN116436828A true CN116436828A (en) 2023-07-14

Family

ID=87080559

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310228544.2A Pending CN116436828A (en) 2023-03-03 2023-03-03 Network traffic prediction and model training method, device, equipment and storage medium

Country Status (1)

Country Link
CN (1) CN116436828A (en)

Similar Documents

Publication Publication Date Title
CN114065863B (en) Federal learning method, apparatus, system, electronic device and storage medium
CN113538235B (en) Training method and device for image processing model, electronic equipment and storage medium
CN114693934B (en) Training method of semantic segmentation model, video semantic segmentation method and device
CN113343803A (en) Model training method, device, equipment and storage medium
CN113780662B (en) Flow prediction method, device, equipment and medium
US20240135698A1 (en) Image classification method, model training method, device, storage medium, and computer program
CN112580733A (en) Method, device and equipment for training classification model and storage medium
CN116489038A (en) Network traffic prediction method, device, equipment and medium
CN115631381A (en) Classification model training method, image classification device and electronic equipment
CN113657468A (en) Pre-training model generation method and device, electronic equipment and storage medium
CN115225543B (en) Flow prediction method and device, electronic equipment and storage medium
CN112396166A (en) Graph convolution neural network training method and device based on mixed granularity aggregator
CN115759232B (en) Method, device, equipment and medium for multitasking parallel processing of deep learning framework
CN113313261B (en) Function processing method and device and electronic equipment
CN116436828A (en) Network traffic prediction and model training method, device, equipment and storage medium
CN113361621B (en) Method and device for training model
CN114758130A (en) Image processing and model training method, device, equipment and storage medium
CN113361574A (en) Training method and device of data processing model, electronic equipment and storage medium
CN115860077B (en) Method, device, equipment and storage medium for processing state data
CN115456167B (en) Lightweight model training method, image processing device and electronic equipment
CN116187473B (en) Federal learning method, apparatus, electronic device, and computer-readable storage medium
CN115018009B (en) Object description method, and network model training method and device
CN110633596A (en) Method and device for predicting vehicle direction angle
CN116502680B (en) Parallel training method and device for mixed expert model
CN114844889B (en) Video processing model updating method and device, electronic equipment and storage medium

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination