Neural network model detection method, apparatus and device, and storage medium

Abstract

The invention provides a neural network model detection method and device, equipment and a storage medium. The method comprises the steps of constructing an attention mechanism diagram of a first neural network model based on structural features of the first neural network model; constructing an adversarial sample function based on the attention mechanism diagram of the first neural network modeland the classification result of the first neural network model for the initial sample; performing updating iteration processing based on the adversarial sample function and the initial sample, and taking an iteration result as an adversarial sample corresponding to the initial sample; and performing classification processing on the adversarial samples through the second neural network model to obtain a classification result of the adversarial samples, and determining the probability of correct classification of the second neural network model according to the difference between the classification result of the adversarial samples and the classification result of the initial samples. According to the invention, accurate adversarial samples can be automatically obtained, and the probabilityof correct classification of the neural network model is improved, so that the robustness detection efficiency is improved.

Description

technical field [0001] The invention relates to artificial intelligence technology, in particular to a detection method, device, electronic equipment and storage medium of a neural network model. Background technique [0002] Artificial Intelligence (AI) is a comprehensive technology of computer science. By studying the design principles and implementation methods of various intelligent machines, the machines have the functions of perception, reasoning and decision-making. Artificial intelligence technology is a comprehensive subject that involves a wide range of fields, such as natural language processing technology and machine learning / deep learning. With the development of technology, artificial intelligence technology will be applied in more fields and play an increasingly important role. increasingly important value. [0003] Neural networks have attracted widespread attention from both academia and industry, and have achieved breakthrough results in several applicatio...

AI Technical Summary

Problems solved by technology

[0004] However, in related technologies, when it is necessary to detect the robustness of the neural network, currently it mainly relies on manual random selection of adversarial samples, and uses the adversarial samples to detect the robustness of the neural network. Low rod accuracy

Images