“This book provides an exceptional foundation for understanding today's large-scale image, video, and text generation systems. Starting from first principles, Yuri builds up to normalizing flows, score matching, and diffusion models. The visualizations bring clarity to concepts that are notoriously difficult to grasp, and the step-by-step derivations serve as an excellent reference.”

“Few books try to connect the mathematical foundations of variational methods with the full landscape of modern generative AI at this level of breadth. The Variational Book does so with clear pedagogical intent, combining step-by-step derivations, conceptual discussion, and illustrations that help make difficult ideas more approachable. For STEM students and technically minded practitioners, this ambitious and unusually wide-ranging treatment offers a substantial guided tour through a fast-moving field.”

“This book has been in the making for years, and that level of care is evident on every page. There are very few resources available today that properly navigate the mathematical foundations of Deep Learning, and Yuri’s work is the first to do so emphasizing the variational aspects. By providing a structured path from foundational concepts to advanced algorithms, it offers a distinct point of view that is both rigorous and highly readable.”

“This book is a great tool to learn the foundations of modern AI methods. It develops the core probability concepts underlying generative state-of-the-art machine learning, and systematically connects them to key techniques such as variational inference, diffusion models, and normalizing flows. ”