A deep learning system and model parameter adjustment method

Abstract

The invention relates to a deep learning system and a method for adjusting model parameters. The system includes: a right-brain-like module is a right-brain-like neural network with a global characteristic memory function; a left-brain-like module is a class-like neural network with a local characteristic response function Left-brain neural network; the similarity filtering module is used to filter the output results of the right-brain-like module by calculating the similarity between the output results of the right-brain-like module and the local regional data, and retain the n outputs with the highest similarity As a result, where n is a natural number greater than 1, the output of the similarity filter module is used as the input of the left-brain module; the game balance module is used to compare the right-brain module and the left-brain module in a game mode including minimax The parameters are adjusted to achieve a game equilibrium between the input of the right-brain-like module and the output of the left-brain-like module. It can resolve the contradiction between global features and local features.

Description

[0001] This application claims the priority of the Chinese patent application submitted to the China Patent Office on December 19, 2017, the application number is 201711378502.8, and the application name is "a deep learning system and network data processing method based on brain-like games", all of which The contents are incorporated by reference in this application. technical field [0002] The invention belongs to the fields of data processing, network security and artificial intelligence, and specifically relates to a deep learning system and a method for adjusting model parameters. Background technique [0003] With the increasing scale of the Internet, a large number of new types of network attack methods have emerged. Facing the current situation that the danger of network attacks is increasing and network security problems are becoming more and more severe, traditional network defense technologies have been difficult to meet the needs of network security. In order to...

AI Technical Summary

Problems solved by technology

[0008] The purpose of the present invention is to solve the above-mentioned problems existing in the existing deep learning system, the present invention provides a deep learning system and a model parameter adjustment method to solve the contradiction between the global feature and the local feature, and realize the integration of the global Intelligent real-time network data processing of and local features, this method is not only applicable to the error back propagation training method of the deep learning system, but also suitable for the non-error back propagation training method of the deep learning system

Images