Overview of Meta-Reinforcement Learning Research
Peng Shengguang
Abstract
Machine learning is a method to achieve artificial intelligence, which is divided into three categories: supervised learning, unsupervised earning, and reinforcement learning. The over-reliance of deep learning on big data restricts its development to some extent, so meta-reinforcement learning (meta-RL) research has received more and more attention. The paper introduces the background, research status and development trend of meta-reinforcement learning, which is also considered as one of the most likely ways to realize general artificial intelligence.
Topics & Concepts
Reinforcement learningMeta learning (computer science)Computer scienceArtificial intelligenceUnsupervised learningLearning classifier systemReinforcementMachine learningHyper-heuristicRobot learningPsychologyEngineeringRobotMobile robotSocial psychologySystems engineeringTask (project management)Reinforcement Learning in RoboticsData Stream Mining TechniquesAdaptive Dynamic Programming Control