No. 118 of 500 No. 9 in Hands-On AI & Programming

Hands-On Machine Learning with Scikit-Learn and PyTorch

Concepts, Tools, and Techniques to Build Intelligent Systems

By Aurélien Géron

Print length
875 pages
Publisher
O'Reilly Media
Reading level
Intermediate
ISBN-13
979-8341607989
Published
December 2, 2025
Written for
Python developers, data analysts and students teaching themselves machine learning and deep learning through code.

Synopsis

This 2025 volume moves Aurélien Géron's Hands-On Machine Learning series from Keras and TensorFlow to PyTorch and the Hugging Face ecosystem. The first half covers classical machine learning with scikit-learn: an end-to-end project running from data exploration to model evaluation, then classification, linear models, decision trees, ensembles, dimensionality reduction and unsupervised techniques such as clustering and anomaly detection. The second half builds neural networks in PyTorch and works through training deep networks, convolutional networks for vision, recurrent models, attention, transformers for language and multimodal tasks, autoencoders, GANs, diffusion models and reinforcement learning.

Géron explains each idea with intuition, diagrams and short code before scaling up, and every chapter has a Jupyter notebook on GitHub that runs in Google Colab. Supplementary tutorials on NumPy, pandas, Matplotlib, linear algebra and calculus help readers fill gaps in their preparation. The later chapters show how to fine-tune pretrained models, including language models, and a chapter on speeding up transformers covers parameter-efficient fine-tuning with LoRA, activation checkpointing, gradient accumulation and parallelism. Readers who finish gain working fluency across the modern stack, from a scikit-learn pipeline to a fine-tuned transformer, and the judgment to diagnose overfitting, tune hyperparameters and choose a method that suits their data.

Scorecard

Editorial score

How it scored

Editorial score: 84.8 / 1005 weighted criteria

  • Influence30%

    24.0/30
  • Authority25%

    21.3/25
  • Practical value20%

    18.0/20
  • Currency15%

    13.5/15
  • Readability10%

    8.0/10
  • Editorial score

    84.8/100

How the rank was computed

Rank No. 118 of 500 comes from a composite of 61.9: observable signals 23.7 of 55 points plus editorial review 38.2 of 45 points.

All 11 inputs
InputPoints
Observable signals23.7/55
Reach (Amazon ratings volume)7.2/16.5
Consensus across best-of lists5.5/13.8
Awards and bestseller evidence0.0/8.3
Reader rating, sample-adjusted6.6/8.3
Amazon sales rank1.6/5.5
Recency2.8/2.8
Editorial review38.2/45
Author authority9.6/11.3
Influence10.8/13.5
Practical value8.1/9.0
Readability3.6/4.5
Currency6.1/6.8

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, code-first path from classical machine learning to transformers and diffusion models, rebuilt around PyTorch for 2025.

Strengths

  • Runnable Colab notebooks for every chapter
  • Spans scikit-learn through diffusion and reinforcement learning
  • Clear explanations with light mathematical demands

Keep in mind

At roughly 880 pages it demands sustained effort, and its mathematical treatment is lighter than the theory-focused textbooks that researchers or graduate students may need.

Key takeaways

  1. Run every project end to end: frame the problem, explore the data, build a pipeline, select and tune a model, then evaluate it on held-out data.
  2. Ensembles such as random forests and boosting often deliver strong results on tabular data, making them sensible baselines before reaching for neural networks.
  3. PyTorch's autograd, modules and explicit training loops give fine control over neural networks, and the same foundations extend to CNNs, RNNs and transformers.
  4. Parameter-efficient fine-tuning such as LoRA, plus gradient accumulation and activation checkpointing, lets practitioners adapt large pretrained transformers on modest hardware.

Notable facts

  • The book's companion notebook repository on GitHub, ageron/handson-mlp, had about 2,300 stars in November 2026. Source (opens in a new tab)

About the author

Aurélien Géron

Aurélien Géron is a machine learning consultant and lecturer. A former Googler, he led YouTube's video classification team from 2013 to 2016. Earlier he was a founder and CTO of Wifirst, a French wireless internet service provider, from 2002 to 2012, and a founder and CTO of the telecom consulting firm Polyconseil. He also wrote the Keras and TensorFlow editions of Hands-On Machine Learning.

Questions

How does this book differ from Hands-On Machine Learning with Scikit-Learn, Keras, and TensorFlow?
The scikit-learn chapters on classical machine learning follow a similar path, but the deep learning half is built on PyTorch and the Hugging Face ecosystem rather than Keras and TensorFlow. It also adds newer material on vision and multimodal transformers, diffusion models, speeding up transformers and fine-tuning pretrained models. Readers committed to TensorFlow may still prefer the earlier edition.
Is Hands-On Machine Learning with Scikit-Learn and PyTorch good for beginners?
It suits beginners who already program in Python and are willing to work through code. Géron explains concepts with intuition and diagrams rather than proofs, and his website offers tutorials on NumPy, pandas, linear algebra and calculus for readers who need a refresher. Complete newcomers to programming should learn Python first, since every chapter is built around runnable notebooks.
Is Hands-On Machine Learning with Scikit-Learn and PyTorch worth reading in 2026?
Yes. Published in October 2025, it reflects the PyTorch-centered tooling most practitioners now use and covers transformers, diffusion models, LoRA-style fine-tuning and reinforcement learning. Its classical machine learning chapters age slowly, and the author maintains companion notebooks on GitHub and an errata page. It remains one of the strongest single books for moving from first models to modern deep learning.

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.