Experience playback sampling reinforcement learning method and system based on confidence upper bound thought

Abstract

The invention provides an experience playback sampling reinforcement learning method and system based on a confidence upper bound thought. The method comprises the steps: collecting the experience obtained through the interaction between an intelligent agent and an environment, and storing the experience data into an experience playback pool; when the current training strategy is updated, experience is randomly selected from the experience playback pool according to the priority probability, and a candidate training sample set is generated; selecting a training sample set according to the confidence upper bound value of each candidate training sample; and updating parameters of a neural network for function approximation according to the training sample data. The technical scheme disclosed by the invention can be combined with any offline RL algorithm, so that the problems of insufficient sample utilization and low learning efficiency of an updating algorithm in related technologies are solved to a certain extent, the sampling efficiency is effectively improved, and the generalization ability of algorithm updating is further improved.

Description

technical field [0001] The disclosure belongs to the technical field of reinforcement learning, and in particular relates to an experience playback sampling reinforcement learning method and system based on the belief upper bound idea. Background technique [0002] The statements in this section merely provide background information related to the present disclosure and do not necessarily constitute prior art. [0003] Deep reinforcement learning is an important research direction in the field of artificial intelligence. Through the process of continuous interaction with the environment, agents autonomously learn the optimal strategy for action execution to maximize their cumulative rewards. Deep reinforcement learning methods have achieved great success in a variety of domains and tasks, including video games, the game of Go, and robot control. Since the huge potential of deep reinforcement learning has not been fully exploited, many works have been devoted to studying its...

AI Technical Summary

Problems solved by technology

[0010] From this point of view, there are many problems in the design principle and generalization ability of the existing sampling technology, which limits its application range.

The method based on experience replay technology still has a lot of room for improvement. Therefore, it is necessary to improve the experience replay sampling method for some problems to improve the sampling efficiency and application potential of deep reinforcement learning algorithms.

Images