Deep neural network interpretation method and device, terminal and storage medium

Abstract

The invention provides a deep neural network interpretation method and device, a terminal and a storage medium. The invention relates to the technical field of artificial intelligence, and the methodspecifically comprises the following steps: determining all target data from a plurality of input data through a result output by a deep neural network; and respectively inputting the target data intoan interpretation factor prediction model to obtain a group of interpretation factors corresponding to each piece of target data, performing statistics on the occurrence frequency of each interpretation factor in the plurality of groups of interpretation factors, performing sorting, and screening a target number of interpretation factors with the highest rank as target interpretation factors of apreset target category. According to the invention, the explanation factor of the prediction result is obtained through the pre-trained explanation factor prediction model, so main factors influencing the deep neural network prediction result are explained; the invention can be applied to smart government affairs, smart city management, smart communities, smart security and protection, smart logistics, smart medical treatment, smart education, smart environmental protection and smart traffic scenes, so that the construction of smart cities is promoted.

Description

technical field [0001] The present invention relates to the technical field of artificial intelligence, in particular to an interpretation method, device, terminal and storage medium of a deep neural network. Background technique [0002] Neural network algorithm is a research hotspot in the field of artificial intelligence since the 1980s. It abstracts the neural network from the perspective of information processing, establishes a simple model, and forms different networks according to different connection methods. It has a self-learning function, which can gradually learn to recognize and predict through training; associative storage function, with high algorithm robustness; high parallelism, with the ability to find optimal solutions at high speed, and can quickly find optimal solutions for complex problems of big data; It has strong plasticity and can fully approach any complex nonlinear relationship; it has strong information synthesis ability and can process quantita...

AI Technical Summary

Problems solved by technology

[0004] The present invention provides an interpretation method, device, terminal and storage medium of a deep neural network to solve the problem that it is difficult to determine factors affecting the prediction results of a deep neural network

Images