Unlock instant, AI-driven research and patent intelligence for your innovation.

A kind of training method and device of graph neural network corresponding to directed graph

A neural network and directed graph technology, applied in the field of graph neural networks, can solve the problems of increasing the data processing burden and increasing the data storage cost of the graph learning system, and achieve the effect of flexible training

Active Publication Date: 2022-07-05
ALIPAY (HANGZHOU) INFORMATION TECH CO LTD
View PDF9 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

In this way, increasing the user's data processing burden also increases the data storage cost of the graph learning system

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • A kind of training method and device of graph neural network corresponding to directed graph
  • A kind of training method and device of graph neural network corresponding to directed graph
  • A kind of training method and device of graph neural network corresponding to directed graph

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0070] The technical solutions of the embodiments of the present specification will be described in detail below with reference to the accompanying drawings.

[0071] The embodiments of this specification disclose a method and device for training a graph neural network corresponding to a directed graph. The following first introduces the application scenarios and technical concepts of the method, as follows:

[0072] As mentioned above, in the current graph learning system, if you want to realize the training of graph neural network based on the same graph data flexibly combined with the direction of the edges, you need to do additional processing on the graph data, such as the DGL system (one A kind of graph learning system), which considers all graphs to be directed graphs. In the DGL system, when a directed graph is used as an undirected graph to train a graph neural network, the directed graph needs to be modified, that is, all the edges of the directed graph need to be mod...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

Embodiments of this specification provide a method and apparatus for training a graph neural network corresponding to a directed graph, the method comprising: controlling a device to obtain a user setting of a first operator for performing node representation aggregation in forward inference computation for the graph neural network After the aggregation configuration information is obtained, it is provided to the working device. The aggregation configuration information includes the participating objects involved in the node representation aggregation and the first edge direction; any first device in the working device is directed to the local graph in the directed graph. The current node is characterized and aggregated by the first operator based on the aggregation configuration information; the current gradient of the graph neural network is determined based on the second operator used for reverse gradient calculation and the gradient source information determined according to the aggregation configuration information , and send it to the control device; the control device updates the model parameters of the graph neural network based on the current gradient sent by each working device.

Description

technical field [0001] The present specification relates to the technical field of graph neural networks, and in particular, to a method and apparatus for training a graph neural network corresponding to a directed graph. Background technique [0002] Graph neural networks are widely used machine learning models. Compared with traditional neural networks, graph neural networks can not only capture the characteristics of nodes, but also describe the characteristics of the relationship between nodes. Therefore, they have achieved excellent results in many machine learning tasks. The relationship network graph corresponding to the graph neural network is divided into directed graph and undirected graph. For example, in a social network, two users are friends, which can be expressed by an undirected graph; in a user-product network, the relationship between users and products Buying behavior can be expressed with a directed graph. [0003] The graph neural network provider can...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Patents(China)
IPC IPC(8): G06N3/08G06K9/62
CPCG06N3/084G06F18/214
Inventor 张国威刘永超
Owner ALIPAY (HANGZHOU) INFORMATION TECH CO LTD