Artificial intelligence reinforcement learning service system

Abstract

The invention discloses an artificial intelligence reinforcement learning service system, and particularly relates to the field of reinforcement learning development platforms, and the system comprises an interface access layer, a service layer, a function layer, a storage layer, an arrangement layer, a model layer and a buffer layer. The interface access layer is an access layer of a cloud service consumer and enables a reinforcement learning researcher to manage a cloud computing environment in a self-service manner; the service layer comprises three parts of a Project-based packaging and management module, a cloud development and debugging environment module and a virtual development environment interface module; the three parts provide various universal or self-defined cloud research environments for field researchers from top to bottom; according to the platform, data, an algorithm and a research environment related to reinforcement learning research are packaged in a virtualization container in a Project form, and an independent test environment is developed for each user using the platform. According to the invention, a visual image of data is provided, code writing and debugging previewing can be carried out at the same time, the effect can be checked in real time, and a researcher can analyze and verify an algorithm.

Description

technical field [0001] The invention relates to the technical field of reinforcement learning development, and more specifically, the invention relates to an artificial intelligence reinforcement learning service system. Background technique [0002] Machine learning is one of the core issues of artificial intelligence, which is to study and simulate human learning behavior, and to generate new knowledge through learning after acquiring knowledge. Data-based machine learning is one of the important methods in modern intelligent technology. Research starts from observed data (samples) to find laws and acquire knowledge, and use these laws and knowledge to predict future data or unobservable data through a certain learning mode. . Machine learning can be classified into supervised learning, unsupervised learning and reinforcement learning according to the learning mode. The goal of reinforcement learning is to learn the mapping from environmental state to behavior, so that t...

AI Technical Summary

Problems solved by technology

[0003] At present, due to the lack of visual development tools on the server side, it is impossible to observe the environment simulation of reinforcement learning agents in real time, and it is difficult to quickly develop and verify codes

Images