Fundamentals of Active Inference

A comprehensive, up-to-date introduction to active inference and the free energy principle for an engineering-focused audience.

This textbook provides comprehensive coverage of the foundational material needed to understand the interdisciplinary, fast-moving field of active inference from first principles. Using a simple, conversational style free of proofs, lemmas, and theorems, Sanjeev Namjoshi brings together theory and technical material in one self-contained text. The book first explains the general statistical framework used in active inference models and then introduces fundamental concepts in machine learning and statistics, connecting them to the active inference perspective. Featuring worked examples, simulations, and detailed walk-throughs of concepts, this user-friendly text aims to expand the readership of active inference to an engineering-focused audience.

• Provides a one-stop-shop for understanding the foundations and applications of active inference and the free energy principle
• Makes a complex, interdisciplinary subject accessible to students and professionals beyond the neurosciences
• Covers discrete and continuous state-space formulations of active inference as well as state-of-the-art extensions to base active inference methods
• Features extensive appendices and supplemental resources

"This is a remarkable book. It is an encyclopedic resource that unpacks everything you always wanted to know about active inference (but were too afraid to ask). It establishes all the foundational concepts and methods that are subsequently applied to furnish an understanding of self-organization through the lens of the free energy principle and active inference. Namjoshi’s forensic, constructive, and didactic approach enables the reader to create a deeply informed and pragmatic account of artificial (and natural) intelligence."
—Karl J. Friston, Professor of Imaging Neuroscience, University College London; coauthor of Active InferenceSanjeev V. Namjoshi is a Machine Learning Engineer at VERSES.AI. Over more than a decade in academia and industry, he performed research in bioinformatics and computational neuroscience as well as lead teams to build machine learning–based software.