Isin solver based on graph convolutional neural network and method for realizing Isin model

Abstract

The invention relates to an Isin solver based on a graph convolutional neural network and a method for realizing an Isin model. The first module in the two core modules is used for the configuration of a graph convolutional network architecture, and the second module is used for the optimization of an equivalent loss function of the Isin Hamiltonian amount. The structure of the graph network is constructed in the first module according to the form of the Isin Hamiltonian quantity. And in the second module, due to the binary property of the Isin Hamiltonian ground state solution and the arbitrariness of the output value of the final convolutional layer, function processing is performed on the output value of the final convolutional layer. An Isin solver using a graph neural network is designed on the basis that a sufficiently strong association action is performed on a GCN structure and a Hamiltonian of quantum annealing, and an NP-hard problem containing large-scale variables can be solved.

Description

technical field [0001] The present invention mainly relates to the technical field of graph convolutional neural network, more precisely, relates to an Ising solver based on graph convolutional neural network and a method for realizing the Ising model. Background technique [0002] Typical combinatorial optimization problems are NP-hard problems, usually including the maximum cut problem (Max-Cut), the maximum independent set problem (MIS), the minimum vertex cover problem, the maximum clique problem and the set cover problem, in biomedicine, financial strategy And traffic planning, circuit design and other aspects have a wide range of applications. The more common solver is the quantum annealer produced by D-wave, but due to the limitation of its hardware structure, this annealer cannot solve large-scale problems, and its calculation speed does not have sufficient advantages compared with simulated annealing. In view of the disadvantages of traditional techniques, it is ve...

AI Technical Summary

Problems solved by technology

For this kind of data with a topological structure, CNN is very difficult to process, and it cannot extract the connection relationship information between nodes, such as judging whether they are connected. This is an important reason for the urgent need to study GCN.

Images