The essential textbooks for ML, deep learning, NLP, and medical AI — with honest notes on who each book is for and where to find free legal copies.
Photo credit: danielfoster437 Library Books via photopin (license)
Updated 2025 — originally published December 2018
A curated reading list of the essential textbooks for ML, NLP, deep learning, and AI research — with honest notes on who each book is for, what it covers, and where to find free legal copies.
The right book at the right stage of your research journey can save you months of confusion. The wrong book — like diving into Murphy’s probabilistic ML before you’re comfortable with basic probability — can discourage you from continuing.
This list is organized by topic and annotated with honest notes. I’ve read or taught from most of these; a few I included based on trusted recommendations. I’ve indicated where free PDFs are legally available.
Table of Contents
- How to Use This List
- Algorithms and Foundations
- Machine Learning (Classical)
- Deep Learning
- Natural Language Processing
- Information Retrieval
- Medical AI and Biostatistics
- Supplementary: Online Courses Worth Treating Like Textbooks
How to Use This List
For a new IIT MS/PhD student: Start with Tom Mitchell’s ML, then Goodfellow’s Deep Learning, then Jurafsky & Martin if you’re doing NLP. Don’t try to read everything — pick one and go deep.
For someone preparing for research interviews: Focus on Murphy (for probabilistic foundations) and the Deep Learning book (for neural network theory).
For someone new to everything: Start with “Hands-On Machine Learning with Scikit-Learn” (Géron) before any of the theory books below. It builds intuition before formalism.
Algorithms and Foundations
Algorithm Design — Kleinberg & Tardos
Who it’s for: Postgrad students in CS taking advanced algorithms courses.
What it covers: Graph algorithms, dynamic programming, network flow, NP-completeness, approximation algorithms.
Free PDF: Yes — widely available online.
My note: Required for IIT CS postgrad coursework. Dense but rigorous.
Networks, Crowds, and Markets — Easley & Kleinberg
Who it’s for: Researchers in social computing, network analysis, computational social science.
What it covers: Game theory, network structure, information cascades, auctions, social networks.
Free PDF: Available at https://www.cs.cornell.edu/home/kleinber/networks-book/
My note: Beautifully written — reads like a story, not a textbook. Essential for anyone working on social media analysis or recommender systems.
Mining of Massive Datasets — Leskovec, Rajaraman, Ullman
Who it’s for: Anyone working with large-scale data.
What it covers: MapReduce, similarity search, link analysis (PageRank), clustering at scale, recommendation systems, stream processing.
Free PDF: Available at http://www.mmds.org/
My note: I bought this book and use it regularly. The chapter on locality-sensitive hashing is one of the clearest explanations I’ve found.
Machine Learning (Classical)
Machine Learning — Tom Mitchell (McGraw Hill)
Who it’s for: Beginners to ML with some math background.
What it covers: Decision trees, neural networks, Bayesian learning, genetic algorithms, reinforcement learning.
Free PDF: Available online.
My note: Start here if you’re new. Concepts are explained lucidly without drowning you in notation.
Pattern Classification — Duda, Hart, Stork
Who it’s for: Graduate students who need a rigorous mathematical foundation.
What it covers: Bayesian decision theory, parametric estimation, discriminant functions, neural networks, feature selection.
Free PDF: Available online.
My note: Contains a fair bit of maths. Recommended for those with a solid probability background. Not a beginner book.
Machine Learning: A Probabilistic Perspective — Kevin P. Murphy
Who it’s for: Advanced PhD students, researchers.
What it covers: Virtually everything in classical ML from a Bayesian perspective — a truly comprehensive reference.
Free PDF: Available online.
My note: This is a reference book, not something to read cover-to-cover. Use it when you need to understand the theory behind a specific algorithm.
Deep Learning
Deep Learning — Goodfellow, Bengio, Courville
Who it’s for: Anyone building neural networks who wants to understand the theory.
What it covers: Linear algebra review, probability, optimization, feedforward networks, CNNs, RNNs, autoencoders, generative models.
Free online version: https://www.deeplearningbook.org/
My note: This is the deep learning textbook. Read Part I (mathematical foundations) even if you skip the rest.
Dive into Deep Learning (d2l.ai) — Zhang et al.
Who it’s for: Practical deep learning learners who want code alongside theory.
What it covers: Everything in Goodfellow, plus transformers, BERT, modern CV, implemented in PyTorch.
Free online version: https://d2l.ai/
My note: Arguably more practical than Goodfellow for 2025. Has interactive code notebooks. This is what I’d recommend to a student starting deep learning today.
Natural Language Processing
Speech and Language Processing — Jurafsky & Martin (3rd edition)
Who it’s for: Anyone doing NLP research.
What it covers: Tokenization, language models, POS tagging, parsing, semantic analysis, transformers, dialogue systems.
Free draft PDF: https://web.stanford.edu/~jurafsky/slp3/
My note: The standard NLP textbook. The 3rd edition draft is regularly updated and includes chapters on transformers and LLMs. Essential.
Information Retrieval
Introduction to Information Retrieval — Manning, Raghavan, Schütze
Who it’s for: Anyone working on search, ranking, or IR systems.
What it covers: Inverted indexes, Boolean retrieval, tf-idf, vector space models, probabilistic retrieval, evaluation.
Free PDF: https://nlp.stanford.edu/IR-book/
My note: Clear, well-organized, and free. Essential for SIGIR/ECIR research.
Medical AI and Biostatistics (2025 Additions)
Biomedical Informatics — Shortliffe & Cimino
Who it’s for: AI researchers entering the clinical domain.
What it covers: Clinical decision support, EHR systems, medical imaging informatics, genomics, ethics.
My note: Foundational text for anyone working in medical AI. Reading even two chapters will help you understand the clinical context your models operate in.
Deep Learning for Medical Image Analysis — Zhou et al.
Who it’s for: Researchers working on radiology, pathology, or any clinical imaging AI.
What it covers: CNNs for medical segmentation, detection, classification; 3D imaging; multi-modal fusion.
My note: A practical, modern reference that bridges computer vision and clinical practice.
Supplementary: Online Courses Worth Treating Like Textbooks
Some online courses are so comprehensive and well-structured that they function as textbooks:
| Course | Platform | Best for |
|---|---|---|
| CS229 Machine Learning | Stanford (free notes) | Rigorous ML theory |
| fast.ai Practical Deep Learning | fast.ai (free) | Hands-on DL intuition |
| AI for Medicine Specialization | Coursera | Clinical AI foundations |
| CS224N NLP with Deep Learning | Stanford (free notes+videos) | Modern NLP |
Final thoughts on machine learning and data science books
We thus learned about the most popular books that will allow you to get started in the exciting domain of machine learning and artificial intelligence. Welcome and all the very best for your journey ahead!
Related articles that will be of interest to you
- How to build an end-to-end Data Science project from scratch
- A Self-help guide to starting your own Machine Learning research project
Discover more from Medical AI Insights
Subscribe to get the latest posts sent to your email.