No. 300 of 500 No. 58 in Machine Learning

Machine Learning

By Tom M. Mitchell

Published
March 1, 1997
Print length
432 pages
Reading level
Intermediate
ISBN-13
978-0070428072
Publisher
McGraw-Hill Education
Written for
Students and engineers who want rigorous, readable grounding in classical machine learning concepts.

Synopsis

Tom Mitchell's 1997 textbook presents machine learning as the study of programs that improve at a task through experience, and it frames each problem by its task, its performance measure and its source of experience. That definition, set out in the first chapter, is still reproduced in course notes today. The thirteen chapters move from concept learning and the general-to-specific ordering of hypotheses through decision trees, artificial neural networks trained by backpropagation, methods for evaluating hypotheses, Bayesian learning, computational learning theory, instance-based methods such as nearest neighbor, genetic algorithms, rule learning, analytical learning, and reinforcement learning.

Mitchell repeatedly asks what inductive bias each method carries and why it generalizes beyond its training data, which gives the book a unifying thread rather than the feel of a catalog. It targets upper-level undergraduates and beginning graduate students and works as a course text, with examples and exercises in each chapter. Readers gain a durable grasp of why learning algorithms work, how to judge whether a learned hypothesis is reliable, and how symbolic and statistical approaches relate. Mitchell offers the full text as a free PDF on his Carnegie Mellon site, along with supplementary chapters on probability estimation, naive Bayes and logistic regression, and key ideas in the field.

Scorecard

Editorial score

How it scored

Editorial score: 76.7 / 1005 weighted criteria

  • Influence30%

    26.4/30
  • Authority25%

    23.3/25
  • Practical value20%

    12.0/20
  • Currency15%

    7.2/15
  • Readability10%

    7.8/10
  • Editorial score

    76.7/100

How the rank was computed

Rank No. 300 of 500 comes from a composite of 52.9: observable signals 18.4 of 55 points plus editorial review 34.5 of 45 points.

All 11 inputs
InputPoints
Observable signals18.4/55
Reach (Amazon ratings volume)7.7/16.5
Consensus across best-of lists5.5/13.8
Awards and bestseller evidence0.0/8.3
Reader rating, sample-adjusted4.8/8.3
Amazon sales rank0.4/5.5
Recency0.0/2.8
Editorial review34.5/45
Author authority10.5/11.3
Influence11.9/13.5
Practical value5.4/9.0
Readability3.5/4.5
Currency3.2/6.8

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

One of the first machine learning textbooks, still clear on fundamentals but silent on the deep learning methods that now dominate.

Strengths

  • Precise definitions that anchor the whole field
  • Clear, well-paced exposition with exercises
  • Free full-text PDF from the author

Keep in mind

Published in 1997, it predates modern deep learning, large-scale ensembles and transformers, so it cannot serve alone as a current course text.

Key takeaways

  1. A program learns when its performance on a task, measured in a defined way, improves with experience, a framing that makes learning problems precise.
  2. Every learning algorithm carries an inductive bias, assumptions beyond the data that let it generalize to examples it has not seen.
  3. Computational learning theory shows how many training examples are needed to learn a hypothesis that is probably approximately correct.
  4. Evaluating hypotheses with confidence intervals and significance tests guards against mistaking chance results on small samples for real accuracy.

Notable facts

  • Mitchell makes the complete book available as a free PDF on his Carnegie Mellon web page, together with three supplementary chapters. Source (opens in a new tab)
  • The book's Carnegie Mellon page links a 2011 machine learning course built on the text, with video lectures, slides, homeworks and exams. Source (opens in a new tab)

About the author

Tom M. Mitchell

Tom M. Mitchell is Founders University Professor at Carnegie Mellon University, where he became the first chair of its Machine Learning Department in 2006 and served as interim dean of the School of Computer Science. He is a member of the National Academy of Engineering, a past president of AAAI, and co-chaired the 2024 National Academies report on AI and the future of work.

Questions

Is Tom Mitchell's Machine Learning still worth reading in 2026?
Yes, for fundamentals. Its definition of learning, its treatment of inductive bias and its chapters on evaluating hypotheses and computational learning theory remain sound and clearly written. It does not cover deep learning, transformers or modern tooling, so pair it with a current deep learning text and a hands-on programming book if you plan to build systems.
Is Mitchell's Machine Learning available for free?
Yes. Tom Mitchell posts the full text of the 1997 book as a free PDF on his Carnegie Mellon University web page, along with errata, related software and data, and three supplementary chapters covering probability estimation with MLE and MAP, naive Bayes and logistic regression, and key ideas in machine learning.
What background do I need for Mitchell's Machine Learning?
The book was written for upper-level undergraduate and introductory graduate courses, so readers should be comfortable with basic probability, logic and algorithms. It does not require prior study of artificial intelligence, and each chapter builds concepts step by step with examples and exercises, which also makes it workable for self-study by motivated engineers.

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.