Image by editor
# introduction
Most of the time, as is often the case with front-end development, you learn better by building something. When I first started coding, I remember spending a month reading about UI/UX, HTML, and CSS, but still couldn’t design a simple interface. Because this type of learning requires practice, projects, and hands-on experience.
Machine learning is different. This field is more rewarding with a deep understanding of theory. It’s not just about applying simple rules like in other fields. If you don’t understand what’s going on under the hood, you’re more likely to hit roadblocks or make mistakes in your model. That’s why I highly recommend reading quality books on machine learning.
This article is part of a new series highlighting books that are free but absolutely worth it. If you are serious about learning and want to strengthen your fundamentals, this list is for you. Let’s start with the first recommendation.
# 1. Understanding Machine Learning: From Theory to Algorithms
Understanding Machine Learning: From Theory to Algorithms introduces machine learning in a rigorous, principled way, starting with the core question of how to transform experience (training data) into expertise (predictive models). Builds from basic theoretical ideas to practical algorithmic paradigms. It provides an extensive discussion of the mathematics behind learning, addressing both the statistical and computational complexity of learning tasks, and covers algorithmic techniques such as stochastic gradient descent, neural networks, and structured output learning, as well as emerging theories such as PAC-bays and compression bounds. Perfect for people who want to go beyond using black-box models and really understand why algorithms behave the way they do.
// Summary overview:
- Fundamentals of Learning (Core Learning Theory, Probably Approximately True (PAC) Learning, Vapnik-Chervonenkis (VC) Dimensions, Generalization, Bias and Complexity Tradeoffs)
- Algorithms and optimization (linear predictors, neural networks, decision trees, boosting, stochastic gradient descent, regularization)
- Model selection and practical considerations (overfitting, underfitting, cross-validation, computational efficiency)
- Unsupervised generative learning (clustering, dimensionality reduction, principal component analysis (PCA), expectation maximization (EM) algorithms, autoencoders)
- Advanced theory and emerging topics (kernel methods, support vector machines (SVM), PAC bays, compression limits, online learning, structured prediction)
# 2. Mathematics for machine learning
Mathematics for Machine Learning bridges the gap between the mathematical foundations and the core techniques of machine learning. It consists of two main parts. The first part covers key mathematical tools such as linear algebra, calculus, probability, and optimization. The second part shows how these tools are used in key machine learning tasks such as regression, classification, density estimation, and dimensionality reduction. While many machine learning books subtitle mathematics, this book focuses on mathematics to help readers actually understand and build machine learning models.
// Summary overview:
- Mathematical foundations for machine learning (linear algebra, analytical geometry, matrix factorization, vector calculus, probability, and continuous optimization)
- Supervised learning and regression (linear regression, Bayesian regression, parameter estimation, empirical risk minimization)
- Dimensionality reduction and unsupervised learning (PCA, Gaussian mixture models, EM algorithms, latent variable modeling)
- Classification and advanced models (SVM, kernels, separating hyperplanes, stochastic modeling, graphical models)
# 3. Overview of statistical learning
Introduction to Statistical Learning (a modern classic in my opinion) is a clear and practical introduction to the field of statistical learning. Statistical learning is essentially a way to use data to make predictions and understand patterns. It covers the key tools you’ll need, including regression, classification, resampling (to check the quality of your model), regularization (to prevent things from going haywire), tree-based methods, SVM, and clustering. We also cover new topics such as deep learning, survival analysis, and processing large numbers of tests at once. Each chapter also includes real-world Python-based labs, so you can not only learn ideas, but also learn how to turn ideas into code.
// Summary overview:
- Fundamentals of statistical learning (overview of statistical learning, supervised and unsupervised learning, regression and classification, model accuracy, trade-offs between bias and variance)
- Linear and nonlinear modeling (linear regression, logistic regression, generalized linear models, polynomial regression, splines, and generalized additive models)
- Advanced prediction methods (tree-based methods, ensemble methods, SVM, deep learning, neural networks)
- Special unsupervised techniques (PCA, clustering, survival analysis, censored data, and multiple testing methods)
# 4. Pattern recognition and machine learning
Pattern recognition and machine learning involves machines learning how to recognize patterns in data. To understand uncertainty, start with the basics of probability and decision making. We then discuss important techniques such as linear regression, classification, neural networks, SVM, and kernel methods. We then discuss more advanced models such as graphical models, mixture models, sampling methods, and sequential models. This book focuses on Bayesian approaches that help you deal with uncertainty and compare models rather than simply finding a single “best” solution. Mathematics may be difficult, but it’s perfect for students and engineers who want a deep understanding of machine learning.
// Summary overview:
- Fundamentals of machine learning (probability theory, Bayesian methods, decision theory, information theory, curse of dimensionality to build a strong conceptual foundation)
- Core models (linear regression and classification, neural networks, kernel methods, sparse models, with emphasis on Bayesian approaches, regularization, and optimization techniques)
- Advanced techniques (graphical models, mixture models using EM, approximate inference, and sampling techniques for complex probabilistic modeling)
- Special topics and applications (continuous latent variable models (PCA, stochastic PCA, kernel PCA), sequential data (hidden Markov models (HMMs), linear dynamical systems (LDS), particle filters), model combination strategies, and a practical appendix on datasets, distributions, and matrix properties.
# 5. Overview of machine learning system
Overview of Machine Learning Systems describes how to build a real machine learning system – not just the model, but the entire setup that makes it work. First, I’ll explain why knowing how to train a model is not enough. You also need to know about data engineering, system design, how hardware and software come together, how to deploy it to the real world, and how to keep things working and secure. We also offer hands-on labs that emphasize the need to think like an engineer and not just a model builder (hardware, resource constraints, pipelines, reliability). The goal is to provide a language, framework, and engineering mindset to move from “I have a model” to “I have a practical AI system that is scalable, robust, and fits real-world needs.”
// Summary overview:
- Fundamentals and design principles (fundamental architecture of machine learning systems, including deployment, machine learning workflows, data engineering, frameworks, and training infrastructure)
- Performance engineering (model optimization, hardware acceleration, inference efficiency, benchmarking, system-level tradeoffs)
- Robust deployment (machine learning operations (MLOps), on-device learning, security and privacy, robustness, and reliability)
- Frontiers of machine learning systems (sustainable AI, AI for good, artificial general intelligence (AGI) systems, new research directions)
# summary
These books cover important parts of machine learning, from mathematics and statistics to real-world systems. Together, they provide a clear path from understanding theory to building and using machine learning models. Which topic should we cover next? Let us know in the comments.
kanwar mereen I’m a machine learning engineer and technical writer with a deep passion for the intersection of data science, AI, and healthcare. She co-authored the e-book “Maximize Productivity with ChatGPT.” She champions diversity and academic excellence as a 2022 Google Generation Scholar for APAC. She has also been recognized as a Teradata Diversity in Tech Scholar, Mitacs Globalink Research Scholar, and a Harvard WeCode Scholar. Kanwal is a passionate advocate for change and founded FEMCodes to empower women in STEM fields.
