Author of 'Probabilistic Machine Learning'
Kevin Murphy
Profile
Kevin Murphy is a Senior Research Scientist at Google DeepMind and the author of what’s become the definitive modern ML reference: the two-volume Probabilistic Machine Learning series published by MIT Press in 2022 and 2023. Over 2,000 pages, built around a unifying Bayesian lens, it covers everything from linear regression to deep generative models to causal inference. If you’re learning ML seriously and need something authoritative, this is what sits on the desk.
Born in Ireland, raised in England, Murphy did his BA at Cambridge, an MEng at the University of Pennsylvania, and a PhD at UC Berkeley under Stuart Russell — the same Russell who co-wrote Artificial Intelligence: A Modern Approach. His 2002 thesis on Dynamic Bayesian Networks is still one of the clearest treatments of the topic. He also built the Bayes Net Toolbox for MATLAB, an open-source package that taught a generation of students how graphical models actually work under the hood.
He spent 2004–2011 as a professor at the University of British Columbia, then moved to Google on sabbatical and never left. His 2012 book Machine Learning: A Probabilistic Perspective won the 2013 DeGroot Prize for best book in statistical science. The 2022/2023 volumes are its rewrite — deep learning had eaten the field in the decade since, and rather than patch the old book, Murphy wrote a new one. He’s also co-editor-in-chief of the Journal of Machine Learning Research.
What makes Murphy useful to developers: he’s one of the few people writing at the reference-textbook level who ships. All the figures in the books are reproducible with accompanying Python code using JAX, PyTorch, and scikit-learn. Draft PDFs are free online. He writes like someone who has to answer questions from both grad students and production engineers — and that’s roughly the audience. Alongside Christopher Bishop’s PRML, Murphy’s books are the two reference works every serious ML practitioner ends up owning.
Books
Probabilistic Machine Learning: An Introduction The 2022 reboot of Murphy's 2012 textbook — a 900+ page foundational survey of modern ML through a Bayesian lens, with runnable Python code for every figure. Probabilistic Machine Learning: Advanced Topics The 2023 sequel — 1,300+ pages on deep generative models, graphical models, Bayesian inference, reinforcement learning, and causality. The most comprehensive advanced ML reference in print. Machine Learning: A Probabilistic Perspective The 2012 original — 1,100 pages that won the DeGroot Prize and defined probabilistic ML pedagogy for a decade. Still useful, though the PML series has largely replaced it.Key Articles & Papers
Probabilistic Machine Learning: An Introduction (free draft) Probabilistic Machine Learning: Advanced Topics (free draft) Dynamic Bayesian Networks: Representation, Inference and Learning The Bayes Net Toolbox for Matlab pyprobml GitHub repositorySpotify Podcasts
