Feature filtering defense method for deep reinforcement learning model

Abstract

The invention discloses a feature filtering defense method for a deep reinforcement learning model. The feature filtering defense method comprises the following steps that: (1) aiming at the]= DDPG model for generating continuous behaviors, a system comprises an actor network and a critic network, wherein the actor network comprises an action estimation network and an action implementation network, the critic network comprises a state estimation network and a state implementation network, pre-training is conducted on the deep reinforcement learning model DDPG, and a current state, a behavior,a reward value and a next state of a pre-training stage are stored in a cache region; (2) an auto-encoder is trained, feature filtering is performed on an input state by using an encoder of the trained auto-encoder to obtain a feature map corresponding to the filtered input state, and a feature map is stored in the cache region; and (3) a convolution kernel in the pre-trained DDPG model is pruned,action prediction is performed by using the pruned DPG model, and a prediction action is output and executed.

Description

technical field [0001] The invention belongs to the technical field of deep learning, and in particular relates to a feature filtering defense method for deep reinforcement learning models. Background technique [0002] With the rapid development of artificial intelligence technology, more and more fields have begun to use AI technology. Since the concept of "artificial intelligence" was first proposed in 1956, AI has attracted more and more attention. His research areas include knowledge representation, machine perception, machine thinking, machine learning, and machine behavior, and he has made some achievements in various fields. Reinforcement learning is also a multi-disciplinary product, which itself is a decision science, so it can be found in many branches of disciplines. Reinforcement learning has a wide range of applications, such as: helicopter aerobatics, game AI, investment management, power station control, making robots imitate human walking, etc. [0003] I...

AI Technical Summary

Problems solved by technology

The existence of singular sample data will increase the training time and may also lead to failure to converge. Therefore, when there is singular sample data, the preprocessing data needs to be normalized before training; otherwise, when there is no singular sample data , then no normalization

Images