Название: Reinforcement Learning Algorithms: Analysis and Applications
Автор: Boris Belousov, Hany Abdulsamad, Pascal Klink, Simone Parisi
Издательство: Springer
Год: 2021
Страниц: 197
Язык: английский
Формат: pdf (true)
Размер: 10.1 MB

This book reviews research developments in diverse areas of reinforcement learning such as model-free actor-critic methods, model-based learning and control, information geometry of policy searches, reward design, and exploration in biology and the behavioral sciences. Special emphasis is placed on advanced ideas, algorithms, methods, and applications.

The contributed papers gathered here grew out of a lecture course on reinforcement learning held by Prof. Jan Peters in the winter semester 2018/2019 at Technische Universitat Darmstadt.

Each chapter of the book provides an overview of a specific topic considered in the lecture, with the parts of the book corresponding to big overarching themes in reinforcement learning. The first part is devoted to the connections with psychology and reward signals in nature. The second part focuses on information-geometric aspects of policy optimization algorithms. The third part covers model-free actor-critic methods, which combine value-based and policy-based algorithms to achieve a better bias-variance trade-off. The fourth part describes model-based approaches, which hold the promise to be more sample-efficient than their model-free counterparts.

The book is intended for machine learning and reinforcement learning students and researchers. Knowledge of linear algebra and statistics is desirable. Nevertheless, all key concepts are introduced in each respective part and chapter of the book, keeping the presentation self-contained and accessible.