Network structure searching method and device

Abstract

The invention discloses a network structure searching method and device, which can improve the network structure searching efficiency. The method comprises the steps that a first total graph is sampled through a first network structure in a network structure search model to generate a plurality of sub-graphs, the sub-graphs are trained at the same time to train the first total graph, the first total graph is constructed according to a search space corresponding to the network structure search model, and the second total graph is constructed according to a search space corresponding to the network structure search model; wherein the search space comprises a plurality of operations, and the first general graph comprises at least one operation in the plurality of operations and a connection line between the operations; wherein the step of training one sub-graph in the plurality of sub-graphs comprises: training the sub-graph by using a batch of training data in a training set; updating the parameters of the first total graph according to the parameters obtained by training the plurality of sub-graphs so as to generate a second total graph; determining a feedback quantity of the firstnetwork structure according to the second total graph; and updating the first network structure according to the feedback quantity of the first network structure.

Description

[0001] Copyright statement [0002] The disclosure of this patent document contains material that is protected by copyright. This copyright belongs to the copyright owner. The copyright owner has no objection to the reproduction by anyone of the patent document or the patent disclosure as it exists in the official records and files of the Patent and Trademark Office. technical field [0003] The present application relates to the field of machine learning, in particular to a network structure search method and device. Background technique [0004] Machine learning algorithms, especially deep learning algorithms, have been developed rapidly and widely used in recent years. As the application scenarios and model structures become more and more complex, it becomes more and more difficult to obtain the optimal model in the application scenarios. Among them, Efficient Neural Architecture Search via Parameter Sharing based on weight sharing can be used. ENAS) to improve the eff...

AI Technical Summary

Problems solved by technology

Although ENAS has improved the speed a lot, because its reinforcement learning is a sequential computing algorithm, the computing efficiency of this single node still has an obvious bottleneck, and the sequential computing will bring a large bias (bias ), affecting the robustness of search results

Images