Neural network architecture search method and device, terminal equipment and storage medium

Abstract

The embodiment of the invention discloses a neural network architecture search method and device, terminal equipment and a storage medium, which are suitable for digital medical treatment. The methodcomprises the following steps: determining a search space and a training data set for constructing a target neural network; adjusting a weight matrix and an architecture parameter in the initial neural network based on the plurality of sample data to obtain an adjusted weight matrix and an adjusted architecture parameter; and determining a target neural network according to the adjusted weight matrix and the basic operation parameters, and determining target input attribute feature data for the target neural network according to the adjusted feature weight parameters. By adopting the embodiment of the invention, the performance of the neural network model can be improved, and the feature selection efficiency of the neural network can be improved.

Description

technical field [0001] The present application relates to the technical field of artificial intelligence, and in particular to a neural network architecture search method, device, terminal equipment and storage medium. Background technique [0002] A neural network with good performance often has an exquisite network structure, which requires a lot of effort to design by highly skilled and experienced human experts. For example, graph neural networks are very popular when analyzing non-Euclidean geometric data, such as social networks, biomedical data, and knowledge graphs, and a lot of research progress has been made using it as a tool. However, at present, the graph neural network does not have good universality, and different graph structure data requires different graph network architectures. However, designing graph neural networks requires a lot of manual work and domain knowledge. In addition, inputting too many attribute features will make the graph neural network ...

AI Technical Summary

Problems solved by technology

However, at present, the graph neural network does not have good universality, and different graph structure data require different graph network architectures

However, designing graph neural networks requires a lot of manual work and domain knowledge

In addition, inputting too many attribute features will make the graph neural network overfit and consume computing resources, and using too few attribute features will make the graph neural network learn poorly.

Images