No. 235 of 500 No. 39 in Machine Learning

Probabilistic Machine Learning

An Introduction

By Kevin P. Murphy

Published
March 1, 2022
Print length
864 pages
Publisher
The MIT Press
Reading level
Advanced
ISBN-13
978-0262046824
Written for
Graduate students, researchers and engineers who want rigorous, probability-based foundations for machine learning.

Synopsis

Kevin Murphy's 2022 textbook replaces his widely used Machine Learning: A Probabilistic Perspective (2012) with an entirely new book that absorbs a decade of deep learning advances. Its organizing idea is that nearly every learning method can be understood as building a probability model and making decisions under uncertainty. Part I lays foundations in probability, statistics, decision theory, information theory, linear algebra and optimization. Later parts move through linear models such as linear and logistic regression; deep neural networks for structured data, images and sequences; nonparametric methods including kernels, trees, random forests and boosting; and learning beyond supervised labels, covering learning with fewer labeled examples, dimensionality reduction, clustering, recommender systems and graph embeddings.

The book is dense and mathematical, written for advanced undergraduates, graduate students and researchers, with end-of-chapter exercises and an appendix on notation. Nearly every figure can be reproduced from companion Python notebooks that use scikit-learn, JAX, PyTorch and TensorFlow, and Murphy hosts a free draft PDF under a Creative Commons license. Readers gain one coherent framework linking classical statistics to modern neural networks, plus the vocabulary needed to read research papers. Generative models such as diffusion, causal inference and reinforcement learning receive fuller treatment in the 2023 sequel, Probabilistic Machine Learning: Advanced Topics.

Scorecard

Editorial score

How it scored

Editorial score: 77.8 / 1005 weighted criteria

  • Influence30%

    25.5/30
  • Authority25%

    22.5/25
  • Practical value20%

    13.0/20
  • Currency15%

    11.3/15
  • Readability10%

    5.5/10
  • Editorial score

    77.8/100

How the rank was computed

Rank No. 235 of 500 comes from a composite of 55.4: observable signals 20.3 of 55 points plus editorial review 35.1 of 45 points.

All 11 inputs
InputPoints
Observable signals20.3/55
Reach (Amazon ratings volume)8.8/16.5
Consensus across best-of lists5.5/13.8
Awards and bestseller evidence0.0/8.3
Reader rating, sample-adjusted4.8/8.3
Amazon sales rank1.2/5.5
Recency0.0/2.8
Editorial review35.1/45
Author authority10.1/11.3
Influence11.5/13.5
Practical value5.9/9.0
Readability2.5/4.5
Currency5.1/6.8

Each criterion is scored from 1 to 10 and weighted, and the points add up to the editorial score. The rank combines observable signals (55 percent) with this score (45 percent). Read the methodology.

Verdict

A comprehensive, rigorous machine learning textbook unified by probability, best suited to readers comfortable with serious mathematics.

Strengths

  • Unified probabilistic treatment of classical and deep methods
  • Companion notebooks reproduce nearly every figure
  • Free draft PDF available from the author

Keep in mind

Its density and mathematical demands make it hard going for self-learners without calculus, linear algebra and probability; it often works better as a reference.

Key takeaways

  1. Treating learning as probabilistic modeling plus Bayesian decision theory unifies regression, classification, neural networks and unsupervised methods in one framework.
  2. Strong foundations in probability, linear algebra and optimization are prerequisites for understanding why modern models work, not optional extras.
  3. Deep neural networks fit naturally within classical statistical ideas such as likelihood, regularization and uncertainty, rather than replacing them.
  4. Reproducible code for nearly every figure lets readers move between equations and working Python implementations in JAX, PyTorch or TensorFlow.

Notable facts

  • Murphy's predecessor book, Machine Learning: A Probabilistic Perspective, won the International Society for Bayesian Analysis's 2013 DeGroot Prize. Source (opens in a new tab)
  • The book's website lists endorsements from Geoffrey Hinton, Christopher Bishop, Daphne Koller, David Blei and Bernhard Schölkopf, among others. Source (opens in a new tab)
  • A free draft PDF is available under a CC BY-NC-ND license, with Colab notebooks that reproduce most figures. Source (opens in a new tab)

About the author

Kevin P. Murphy

Kevin P. Murphy is a principal research scientist and director at Google DeepMind and an adjunct professor at the University of British Columbia, where he was associate professor of computer science and statistics from 2004 to 2012. He holds a PhD from UC Berkeley, and his 2012 textbook Machine Learning: A Probabilistic Perspective won the DeGroot Prize.

Questions

Is Probabilistic Machine Learning: An Introduction worth reading in 2026?
Yes, for readers who want deep foundations. Its material on probability, statistics, optimization and neural networks has not gone out of date, and the companion code and a free draft PDF remain available online. It predates the large language model boom, so readers focused on transformers at scale, fine-tuning or AI agents will need other sources, including Murphy's 2023 Advanced Topics volume.
Is Probabilistic Machine Learning good for beginners?
Not for true beginners. Although titled an introduction, it assumes comfort with calculus, linear algebra and probability, and it moves quickly through demanding derivations. Motivated newcomers can use Part I as a refresher, but most will benefit from a gentler text first. Graduate students, researchers and engineers who already know basic machine learning get the most from it.
What is the difference between Murphy's Introduction and Advanced Topics books?
The Introduction (2022) covers foundations, linear models, deep neural networks, nonparametric methods and learning beyond supervised labels. The Advanced Topics volume (2023) extends the same probabilistic approach to subjects such as deep generative models, Bayesian inference, causality and reinforcement learning. Both grew out of Murphy's 2012 Machine Learning: A Probabilistic Perspective, and the Introduction is the natural starting point.

Sources

As an Amazon Associate, Best AI Books earns from qualifying purchases. Best AI Books is sponsored by Iternal.ai, whose founder and CEO, John Byron Hanby IV, wrote The AI Strategy Blueprint, our #1 ranked book. How we rank books.