Hyper-network optimization method and system based on tensor

Abstract

The invention provides a tensor-based super network optimization method and system. The method comprises the following steps: constructing a tensor network corresponding to a super network; determining an optimal sub-network having an optimal structure among at least one sub-network of the super-network based on the constructed tensor network; training the learning model of the optimal sub-network; and obtaining an embedded representation of the optimal sub-network by using the trained learning model.

Description

technical field [0001] The present application relates to graph network technology in the field of artificial intelligence, and more particularly, to a tensor-based hypernetwork optimization method and system. Background technique [0002] With the rapid development of information network technology, various information network data content presents an explosive growth trend. Such content is usually large-scale, heterogeneous, and loosely organized, posing challenges for people to effectively acquire information and knowledge. [0003] In real-life and scientific research fields, graphs are usually used to represent data. For example, graphs are used to represent transportation networks, graphs to represent social networks, graphs to represent knowledge graphs, and graphs to represent molecular structures. A graph may include graph nodes and edges, where an edge represents a relationship between two or more graph nodes. [0004] A supernet is a special kind of graph netwo...

AI Technical Summary

Problems solved by technology

Furthermore, the state-of-the-art solutions are inefficient since the topology of the hypernetwork is highly task-dependent and diverse.

Images