This book becomes useful the moment you need to bring math, data, and code together instead of treating them as three separate topics.
The chapter moves from basic algorithms and data analysis into first hands-on experiments in TensorFlow, Keras, and PyTorch, where the point is not the library names but the habit of running models and comparing results.
For interviews and architecture conversations, it works as a gentle entry point that ties machine learning fundamentals to how engineers actually learn to work with data, models, and experiments.
Practical value of this chapter
Deep learning base
The chapter quickly turns math, classical machine learning algorithms, and first deep learning ideas into one coherent map.
Tooling practice
TensorFlow, Keras, and PyTorch make it easy to see how the same path from idea to experiment looks across different toolchains.
Bridge to modern AI
Once this base is in place, it becomes much easier to move into large language models and AI engineering without treating modern systems as just a list of new names.
Interview material
The book helps you explain not only the terminology, but also how an engineer moves from theory to first experiments and working solutions.
Source
Telegram: Book Cube
A detailed review with chapter-by-chapter notes and a practical reading perspective.
Глубокое обучение и анализ данных. Практическое руководство
Authors: Дмитрий Малов
Publisher: БХВ-Петербург, 2023
Length: 270 pages
A concise and practical introduction to deep learning and data analysis: fundamentals, classical machine learning algorithms, applied tasks, and hands-on work with TensorFlow, Keras, and PyTorch.
What this book covers
The book closes a common gap: math and classical machine learning algorithms are usually taught apart from applied code, and turning them into a working skill is hard. Here the path runs straight through — mathematical foundations and classical algorithms first, then deep learning and hands-on work in three popular frameworks.
The price of that pace is breadth over depth: the move from classical ML models to neural networks is fast, and backpropagation and the first experiments are covered just enough to run the code, not to derive every formula step by step.
Structure: 8 chapters
Machine learning foundations
Linear algebra, probability basics, and core machine learning task framing: classification, regression, anomaly detection, translation, and synthesis. Includes a short introduction to Python, OOP, and the development process.
Core machine learning algorithms
Data preprocessing, dimensionality reduction, linear and logistic regression, decision trees, support vector machines, naive Bayes, k-means, kNN, random forest, and gradient boosting.
Deep learning foundations
Backpropagation, perceptron, Markov chains, Boltzmann machines, Hopfield networks, convolutional and recurrent networks, transformers, autoencoders, and GANs.
Data science foundations
CRISP-DM methodology, key roles in an ML team, trends such as deepfakes, AutoML, and MLOps, plus an overview of TensorFlow, PyTorch, and Keras.
Deep learning tasks
Hands-on examples: data augmentation, computer vision with OpenCV, symbol recognition, text processing, audio processing, and video processing.
TensorFlow
Applying chapter 5 tasks using TensorFlow.
Keras
Rebuilding practical tasks through the higher-level Keras API.
PyTorch
Implementing the same class of tasks in the PyTorch ecosystem.
What is most useful
- The intro is short and coherent: you cover the path from math to applied tasks in one pass, without a long runway.
- Working code ships with it — you can take it as a starting point and reshape it for your own experiments instead of writing from scratch.
- The book closes the gap between introductory theory and a first practical project, which is exactly where beginners stall.
What to keep in mind
- Coverage is broad, and the cost is depth: many topics stay high-level, so you will go elsewhere for the details.
- Chapter 2 is dense with algorithms per page — a beginner can lose the context and skim past the point.
- Without separate sources on math and machine learning you probably won't reach real practical confidence; one book is not enough.
Practice
GitHub repository
Code and charts from the book that make it easy to replay the examples and build your own variations.
How to get maximum value from it
How you read it matters: move quickly through the foundational chapters and practice alongside them. Topics like backpropagation barely settle on theory alone — they are easier to hold once you see them running in code.
Take chapters 1-4 first — they align the vocabulary and fundamentals you need before the rest stops making sense.
On chapters 5-8 don't read the code, run it from the repository and change parameters, data, and architectures by hand: theory sticks faster that way.
After the book, move on to resources about real AI systems, large language models, and modern AI engineering — where quality, cost, and operations enter the picture.
Related chapters
- Why engineers should know ML and AI - Provides the AI/ML section map and shows where this book fits in the broader learning path.
- Precision and Recall Basics - Without these two metrics you cannot honestly judge the book's classification tasks: this explains what they actually measure and where they get confused.
- AI Engineering (short summary) - Extends the path from foundational ML practice to engineering AI systems where quality, cost, and operations matter.
- Hands-On Large Language Models (short summary) - Bridges from classical deep learning to large language models, vector representations, and RAG architectures.
- Prompt Engineering for LLMs (short summary) - Extends the applied track with prompt and context design practices for LLM applications.