Hands-On Machine Learning with Scikit-Learn and PyTorch. Concepts, Tools, and Techniques to Build Intelligent Systems (ebook) Koziegłowy

The potential of machine learning today is extraordinary, yet many aspiring developers and tech professionals find themselves daunted by its complexity. Whether youre looking to enhance your skill set and apply machine learning to real-world projects or are simply curious about how AI systems function, this book is your jumping-off place. With an approachable yet deeply informative style, author Aurélien Géron delivers the ultimate introductory guide to machine learning and deep learning. Drawing on the Hugging Face ecosystem, with a focus on clear explanations and real-world examples, the book takes you through cutting-edge tools like Scikit-Learn and PyTorch—from basic regression techniques to advanced neural networks. Whether youre a student, professional, or hobbyist, youll gain the skills to build intelligent systems. Understand ML basics, including concepts like overfitting and hyperparameter tuning Complete an end-to-end ML project using scikit-Learn, covering everything from data exploration to model evaluation Learn techniques for unsupervised learning, such as clustering and anomaly detection Build advanced architectures like transformers and diffusion models with PyTorch Harness the power of pretrained models—including LLMs—and learn to fine-tune them Train autonomous agents using reinforcement learning Spis treści: Preface Machine Learning in Your Projects Objective and Approach Code Examples Prerequisites Roadmap Changes Between the TensorFlow and PyTorch Versions Other Resources Conventions Used in This Book OReilly Online Learning How to Contact Us Acknowledgments I. The Fundamentals of Machine Learning 1. The Machine Learning Landscape What Is Machine Learning? Why Use Machine Learning? Examples of Applications Types of Machine Learning Systems Training Supervision Supervised learning Unsupervised learning Semi-supervised learning Self-supervised learning Reinforcement learning Batch Versus Online Learning Batch learning Online learning Instance-Based Versus Model-Based Learning Instance-based learning Model-based learning and a typical machine learning workflow Main Challenges of Machine Learning Insufficient Quantity of Training Data Nonrepresentative Training Data Poor-Quality Data Irrelevant Features Overfitting the Training Data Underfitting the Training Data Deployment Issues Stepping Back Testing and Validating Hyperparameter Tuning and Model Selection Data Mismatch Exercises 2. End-to-End Machine Learning Project Working with Real Data Look at the Big Picture Frame the Problem Select a Performance Measure Check the Assumptions Get the Data Running the Code Examples Using Google Colab Saving Your Code Changes and Your Data The Power and Danger of Interactivity Book Code Versus Notebook Code Download the Data Take a Quick Look at the Data Structure Create a Test Set Explore and Visualize the Data to Gain Insights Visualizing Geographical Data Look for Correlations Experiment with Attribute Combinations Prepare the Data for Machine Learning Algorithms Clean the Data Handling Text and Categorical Attributes Feature Scaling and Transformation Custom Transformers Transformation Pipelines Select and Train a Model Train and Evaluate on the Training Set Better Evaluation Using Cross-Validation Fine-Tune Your Model Grid Search Randomized Search Ensemble Methods Analyzing the Best Models and Their Errors Evaluate Your System on the Test Set Launch, Monitor, and Maintain Your System Try It Out! Exercises 3. Classification MNIST Training a Binary Classifier Performance Measures Measuring Accuracy Using Cross-Validation Confusion Matrices Precision and Recall The Precision/Recall Trade-Off The ROC Curve Multiclass Classification Error Analysis Multilabel Classification Multioutput Classification Exercises 4. Training Models Linear Regression The Normal Equation Computational Complexity Gradient Descent Batch Gradient Descent Stochastic Gradient Descent Mini-Batch Gradient Descent Polynomial Regression Learning Curves Regularized Linear Models Ridge Regression Lasso Regression Elastic Net Regression Early Stopping Logistic Regression Estimating Probabilities Training and Cost Function Decision Boundaries Softmax Regression Exercises 5. Decision Trees Training and Visualizing a Decision Tree Making Predictions Estimating Class Probabilities The CART Training Algorithm Computational Complexity Gini Impurity or Entropy? Regularization Hyperparameters Regression Sensitivity to Axis Orientation Decision Trees Have a High Variance Exercises 6. Ensemble Learning and Random Forests Voting Classifiers Bagging and Pasting Bagging and Pasting in Scikit-Learn Out-of-Bag Evaluation Random Patches and Random Subspaces Random Forests Extra-Trees Feature Importance Boosting AdaBoost Gradient Boosting Histogram-Based Gradient Boosting Stacking Exercises 7. Dimensionality Reduction The Curse of Dimensionality Main Approaches for Dimensionality Reduction Projection Manifold Learning PCA Preserving the Variance Principal Components Projecting Down to d Dimensions Using Scikit-Learn Explained Variance Ratio Choosing the Right Number of Dimensions PCA for Compression Randomized PCA Incremental PCA Random Projection LLE Other Dimensionality Reduction Techniques Exercises 8. Unsupervised Learning Techniques Clustering Algorithms: k-means and DBSCAN k-Means Clustering The k-means algorithm Centroid initialization methods Accelerated k-means and mini-batch k-means Finding the optimal number of clusters Limits of k-Means Using Clustering for Image Segmentation Using Clustering for Semi-Supervised Learning DBSCAN Other Clustering Algorithms Gaussian Mixtures Using Gaussian Mixtures for Anomaly Detection Selecting the Number of Clusters Bayesian Gaussian Mixture Models Other Algorithms for Anomaly and Novelty Detection Exercises II. Neural Networks and Deep Learning 9. Introduction to Artificial Neural Networks From Biological to Artificial Neurons Biological Neurons Logical Computations with Neurons The Perceptron The Multilayer Perceptron and Backpropagation Building and Training MLPs with Scikit-Learn Regression MLPs Classification MLPs Hyperparameter Tuning Guidelines Number of Hidden Layers Number of Neurons per Hidden Layer Learning Rate Batch Size Other Hyperparameters Exercises 10. Building Neural Networks with PyTorch PyTorch Fundamentals PyTorch Tensors Hardware Acceleration Autograd Implementing Linear Regression Linear Regression Using Tensors and Autograd Linear Regression Using PyTorchs High-Level API Implementing a Regression MLP Implementing Mini-Batch Gradient Descent Using DataLoaders Model Evaluation Building Nonsequential Models Using Custom Modules Building Models with Multiple Inputs Building Models with Multiple Outputs Building an Image Classifier with PyTorch Using TorchVision to Load the Dataset Building the Classifier Fine-Tuning Neural Network Hyperparameters with Optuna Saving and Loading PyTorch Models Compiling and Optimizing a PyTorch Model Exercises 11. Training Deep Neural Networks The Vanishing/Exploding Gradients Problems Glorot Initialization and He Initialization Better Activation Functions Leaky ReLU ELU and SELU GELU, Swish, SwiGLU, Mish, and RELU2 Batch Normalization Implementing batch norm with PyTorch Batch norm 1D, 2D, and 3D Layer Normalization Gradient Clipping Reusing Pretrained Layers Transfer Learning with PyTorch Unsupervised Pretraining Pretraining on an Auxiliary Task Faster Optimizers Momentum Nesterov Accelerated Gradient AdaGrad RMSProp Adam AdaMax NAdam AdamW Learning Rate Scheduling Exponential Scheduling Cosine Annealing Performance Scheduling Warming Up the Learning Rate Cosine Annealing with Warm Restarts 1cycle Scheduling Avoiding Overfitting Through Regularization 1 and 2 Regularization Dropout Monte Carlo Dropout Max-Norm Regularization Practical Guidelines Exercises 12. Deep Computer Vision Using Convolutional Neural Networks The Architecture of the Visual Cortex Convolutional Layers Filters Stacking Multiple Feature Maps Implementing Convolutional Layers with PyTorch Pooling Layers Implementing Pooling Layers with PyTorch CNN Architectures LeNet-5 AlexNet GoogLeNet ResNet Xception SENet Other Noteworthy Architectures Choosing the Right CNN Architecture GPU RAM Requirements: Inference Versus Training Reversible Residual Networks (RevNets) Implementing a ResNet-34 CNN Using PyTorch Using TorchVisions Pretrained Models Pretrained Models for Transfer Learning Classification and Localization Object Detection Fully Convolutional Networks You Only Look Once Object Tracking Semantic Segmentation Exercises 13. Processing Sequences Using RNNs and CNNs Recurrent Neurons and Layers Memory Cells Input and Output Sequences Training RNNs Forecasting a Time Series The ARMA Model Family Preparing the Data for Machine Learning Models Forecasting Using a Linear Model Forecasting Using a Simple RNN Forecasting Using a Deep RNN Forecasting Multivariate Time Series Forecasting Several Time Steps Ahead Forecasting Using a Sequence-to-Sequence Model Handling Long Sequences Fighting the Unstable Gradients Problem Tackling the Short-Term Memory Problem LSTM cells GRU cells Using 1D convolutional layers to process sequences WaveNet Exercises 14. Natural Language Processing with RNNs and Attention Generating Shakespearean Text Using a Character RNN Creating the Training Dataset Embeddings Building and Training the Char-RNN Model Generating Fake Shakespearean Text Sentiment Analysis Using Hugging Face Libraries Tokenization Using the Hugging Face Tokenizers Library Reusing Pretrained Tokenizers Building and Training a Sentiment Analysis Model Bidirectional RNNs Reusing Pretrained Embeddings and Language Models Task-Specific Classes The Trainer API Hugging Face Pipelines An Encoder-Decoder Network for Neural Machine Translation Beam Search Attention Mechanisms Exercises 15. Transformers for Natural Language Processing and Chatbots Attention Is All You Need: The Original Transformer Architecture Positional Encodings Multi-Head Attention Building the Rest of the Transformer Building an English-to-Spanish Transformer Encoder-Only Transformers for Natural Language Understanding BERTs Architecture BERT Pretraining BERT Fine-Tuning Other Encoder-Only Models RoBERTa by Facebook AI, July 2019 (125M to 355M parameters) DistilBERT by Hugging Face, October 2019 (66M) ALBERT by Google Research, December 2019 (12M235M) ELECTRA by Google Research, March 2020 (14M335M) DeBERTa by Microsoft, January 2021 (139M1.5B) More encoder-only models on Hugging Face Hub Decoder-Only Transformers GPT-1 Architecture and Generative Pretraining GPT-2 and Zero-Shot Learning GPT-3, In-Context Learning, One-Shot Learning, and Few-Shot Learning Using GPT-2 to Generate Text Using GPT-2 for Question Answering Downloading and Running an Even Larger Model: Mistral-7B Turning a Large Language Model into a Chatbot Fine-Tuning a Model for Chatting and Following Instructions Using SFT and RLHF Direct Preference Optimization (DPO) Fine-Tuning a Model Using the TRL Library From a Chatbot Model to a Full Chatbot System Model Context Protocol Libraries and Tools Encoder-Decoder Models Exercises 16. Vision and Multimodal Transformers Vision Transformers RNNs with Visual Attention DETR: A CNN-Transformer Hybrid for Object Detection The Original ViT Implementing a ViT from scratch using PyTorch Fine-tuning a pretrained ViT using the Transformers library Data-Efficient Image Transformer Pyramid Vision Transformer for Dense Prediction Tasks The Swin Transformer: A Fast and Versatile ViT DINO: Self-Supervised Visual Representation Learning Other Major Vision Models and Techniques Multimodal Transformers VideoBERT: A BERT Variant for Text plus Video ViLBERT: A Dual-Stream Transformer for Text plus Image CLIP: A Dual-Encoder Text plus Image Model Trained with Contrastive Pretraining DALLE: Generating Images from Text Prompts Perceiver: Bridging High-Resolution Modalities with Latent Spaces Perceiver IO: A Flexible Output Mechanism for the Perceiver Flamingo: Open-Ended Visual Dialogue BLIP and BLIP-2 Other Multimodal Models Exercises 17. Speeding Up Transformers 18. Autoencoders, GANs, and Diffusion Models Efficient Data Representations Performing PCA with an Undercomplete Linear Autoencoder Stacked Autoencoders Implementing a Stacked Autoencoder Using PyTorch Visualizing the Reconstructions Anomaly Detection Using Autoencoders Visualizing the Fashion MNIST Dataset Unsupervised Pretraining Using Stacked Autoencoders Tying Weights Training One Autoencoder at a Time Convolutional Autoencoders Denoising Autoencoders Sparse Autoencoders Variational Autoencoders Generating Fashion MNIST Images Discrete Variational Autoencoders Generative Adversarial Networks The Difficulties of Training GANs Diffusion Models Exercises 19. Reinforcement Learning What Is Reinforcement Learning? Policy Gradients Introduction to the Gymnasium Library Neural Network Policies Evaluating Actions: The Credit Assignment Problem Solving the CartPole Using Policy Gradients Value-Based Methods Markov Decision Processes Temporal Difference Learning Q-Learning Exploration Policies Approximate Q-Learning and Deep Q-Learning Implementing Deep Q-Learning DQN Improvements Actor-Critic Algorithms Mastering Atari Breakout Using the Stable-Baselines3 PPO Implementation Overview of Some Popular RL Algorithms Exercises Thank You! A. Autodiff Manual Differentiation Finite Difference Approximation Forward-Mode Autodiff Reverse-Mode Autodiff B. Mixed Precision and Quantization Common Number Representations Reduced Precision Models Mixed-Precision Training Quantization Linear Quantization Post-Training Quantization Using torch.ao.quantization Dynamic quantization Static quantization Quantization-Aware Training (QAT) Quantizing LLMs Using the bitsandbytes Library Using Pre-Quantized Models Index O autorze: Aurélien Géron - jest konsultantem do spraw uczenia maszynowego. Wcześniej pracował w korporacji Google, a w latach 2013 – 2016 kierował zespołem klasyfikowania filmów w firmie YouTube. Był również założycielem i dyrektorem do spraw technicznych (w latach 2002 – 2012) w firmie Wifirst — czołowym francuskim dostawcy bezprzewodowych usług internetowych; te same funkcje pełnił w 2001 roku w firmie Polyconseil — obecnie zarządza ona usługą udostępniania samochodów elektrycznych Autolib’.