Algorithms for Decision Making (algorithmsbook.com)

A new textbook, Algorithms for Decision Making, is available as a freely downloadable PDF (Creative Commons CC-BY-NC-ND) with exclusive copyright licensed to The MIT Press; individual chapters can also be downloaded and the PDF is kept current via GitHub/errata. The book presents a broad, unified introduction to decision making under uncertainty, making it a practical resource for researchers, students, and engineers working in AI/ML, robotics, and reinforcement learning who need both foundational theory and algorithmic recipes. Technically, the book spans representation and inference through parameter and structure learning; exact and approximate solution methods for planning; model-based and model-free approaches; policy search and policy-gradient techniques (including actor–critic methods and gradient estimation/optimization); exploration–exploitation tradeoffs; imitation learning; belief-state planning (exact, offline, online); controller abstractions; and multiagent reasoning. Useful appendices cover mathematical tools, probability distributions, computational complexity, neural representations, search algorithms, problem sets, and Julia examples, making it suitable for implementation and classroom use. Overall, the book synthesizes theoretical foundations with practical algorithms, serving as a go-to reference for anyone designing or analyzing decision systems under uncertainty.