Network structure search method and device, computer storage medium and computer program product

Abstract

A network structure search method include the following steps: a search space define step: (S12), determining the search space of a neural network model to be subjected to network structure search, wherein the search space defines a plurality of operations on an operation layer between every two nodes in the neural network model; a pre-training step: (S14), training a total graph of the search space with preset parameters of a first network structure according to the first network structure to obtain a total graph with pre-training parameters, wherein the total graph is composed of operations;and a training step: (S16), training the total graph with the pre-training parameters according to the first network structure and updating the first network structure according to the feedback quantity of the first network structure. The invention further discloses a network structure searching device, a computer storage medium and a computer program product.

Description

technical field [0001] The present application relates to the field of machine learning, and in particular to a method and device for searching a network structure, a computer storage medium and a computer program product. Background technique [0002] In related technologies, machine learning algorithms, especially deep learning algorithms, have been rapidly developed 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 efficiency of Neural Architecture Search (NAS). However, the network structure searched by ENAS often has a large bias (bias), that is, the network structure searched by ENAS always tends to operate with a larger convolution kernel size (kernel size). This leads to large paramete...

AI Technical Summary

Problems solved by technology

However, the network structure searched by ENAS often has a large bias (bias), that is, the network structure searched by ENAS will always tend to operate with a larger convolution kernel size (kernel size)

This leads to large parameters of the searched model, which is not easy to debug and train

In addition, the bias of the controller means that the local optimal solution to which the controller converges cannot fully explore (explore) the search space

Such a controller does not have high credibility, and cannot guarantee that the searched model is the global optimal

Images