Review of learning environments for AI agents

Author: Zurabi Isakadze
Co-authors: Zurabi Isakadze
Keywords: Reinforcement Learning, Learning Environments, Artificial Intelligence
Annotation:

A learning environment (LE) or a world in which an agent operates, is a crucial part of the given task. The type of actions an agent can make and information it receives from the environment is directly related to the algorithms that can be effective for solving the problem. Computer games naturally satisfy many of the requirements of LE, hence In the last years they have been used for testing and benchmarking Reinforcement Learning algorithms. Besides, some of the novel environments have been created as well. In this seminar we review several such Artificial Learning Environments that are heavily used in the research field, best results that have been achieved on them and discuss what types of LEs might be interesting when working on AGI systems.