Probabilistic Machine Learning: Advanced Topics

Livro digital

Título:
Probabilistic Machine Learning: Advanced Topics

Autor:
Kevin P. Murphy

Categoria:
Tecnologia > IA

Doador:
Raffaello D. N.

Sinopse:
Most machine learning textbooks stop where the hard questions begin. Probabilistic Machine Learning: Advanced Topics starts exactly there — with the mathematical substrate that serious practitioners need but rarely find unified in one place. The book is organized around six thematic arcs: Fundamentals, Inference, Prediction, Generation, Discovery, and Action, building a rigorous framework that connects classical Bayesian theory to modern deep learning architectures. The depth across each part is deliberate. Inference alone spans seven chapters covering Gaussian filtering, message passing, variational methods, and three distinct approaches to Monte Carlo — tools rarely gathered in a single volume. Prediction includes Bayesian neural networks and Gaussian processes, offering a probabilistic lens on deep learning that most courses omit. Part IV on Generation is where the book separates itself: variational autoencoders, normalizing flows, energy-based models, diffusion models, and GANs are all treated within a unified probabilistic framework — one of the few texts that explains why diffusion models work, not merely how. The final two parts cover representation learning, interpretability, reinforcement learning, and a dedicated chapter on causality. At 1,370 pages, this is not a book one reads — it is a book one works through. Kevin P. Murphy writes for practitioners and researchers who want to reason about uncertainty rather than just optimize for accuracy. Published by MIT Press and freely available under a Creative Commons license, it has already shaped graduate curricula across top institutions. Readers who complete it leave with something rare: a coherent mathematical vocabulary for the entire modern machine learning landscape.