book

Deep Learning with Python, Third Edition

Name: Deep Learning with Python, Third Edition
ISBN: 9781633436589

by Matthew Watson, Francois Chollet

September 2025

Intermediate to advanced

648 pages

20h 1m

English

Manning Publications

Audiobook available

Read now

Unlock full access

Deep Learning with Python, Third Edition
copyright
contents
Praise for the Second Editio
preface
acknowledgments
about this book
about the authors
about the cover illustration
1 What is deep learning?
1.1 Artificial intelligence, machine learning, and deep learning1.2 Artificial intelligence1.3 Machine learning1.4 Learning rules and representations from data1.5 The “deep” in “deep learning”1.6 Understanding how deep learning works, in three figures1.7 What makes deep learning different1.8 The age of generative AI1.9 What deep learning has achieved so far1.10 Beware of the short-term hype1.11 Summer can turn to winter1.12 The promise of AI

2 The mathematical building blocks of neural networks
2.1 A first look at a neural network2.2 Data representations for neural networks2.2.1 Scalars (rank-0 tensors)2.2.2 Vectors (rank-1 tensors)2.2.3 Matrices (rank-2 tensors)2.2.4 Rank-3 tensors and higher-rank tensors2.2.5 Key attributes2.2.6 Manipulating tensors in NumPy2.2.7 The notion of data batches2.2.8 Real-world examples of data tensors2.3 The gears of neural networks: Tensor operations2.3.1 Element-wise operations2.3.2 Broadcasting2.3.3 Tensor product2.3.4 Tensor reshaping2.3.5 Geometric interpretation of tensor operations2.3.6 A geometric interpretation of deep learning2.4 The engine of neural networks: Gradient-based optimization2.4.1 What’s a derivative?2.4.2 Derivative of a tensor operation: The gradient2.4.3 Stochastic gradient descent2.4.4 Chaining derivatives: The Backpropagation algorithm2.5 Looking back at our first example2.5.1 Reimplementing our first example from scratch2.5.2 Running one training step2.5.3 The full training loop2.5.4 Evaluating the model2.6 Summary
3 Introduction to TensorFlow, PyTorch, JAX, and Keras
3.1 A brief history of deep learning frameworks3.2 How these frameworks relate to each other3.3 Introduction to TensorFlow3.3.1 First steps with TensorFlow3.3.2 An end-to-end example: A linear classifier in pure TensorFlow3.3.3 What makes the TensorFlow approach unique3.4 Introduction to PyTorch3.4.1 First steps with PyTorch3.4.2 An end-to-end example: A linear classifier in pure PyTorch3.4.3 What makes the PyTorch approach unique3.5 Introduction to JAX3.5.1 First steps with JAX3.5.2 Tensors in JAX3.5.3 Random number generation in JAX3.5.4 An end-to-end example: A linear classifier in pure JAX3.5.5 What makes the JAX approach unique3.6 Introduction to Keras3.6.1 First steps with Keras3.6.2 Layers: The building blocks of deep learning3.6.3 From layers to models3.6.4 The “compile” step: Configuring the learning process3.6.5 Picking a loss function3.6.6 Understanding the fit method3.6.7 Monitoring loss and metrics on validation data3.6.8 Inference: Using a model after training
4 Classification and regression
4.1 Classifying movie reviews: A binary classification example4.1.1 The IMDb dataset4.1.2 Preparing the data4.1.3 Building your model4.1.4 Validating your approach4.1.5 Using a trained model to generate predictions on new data4.1.6 Further experiments4.1.7 Wrapping up4.2 Classifying newswires: A multiclass classification example4.2.1 The Reuters dataset4.2.2 Preparing the data4.2.3 Building your model4.2.4 Validating your approach4.2.5 Generating predictions on new data4.2.6 A different way to handle the labels and the loss4.2.7 The importance of having sufficiently large intermediate layers4.2.8 Further experiments4.2.9 Wrapping up4.3 Predicting house prices: A regression example4.3.1 The California Housing Price dataset4.3.2 Preparing the data4.3.3 Building your model4.3.4 Validating your approach using K-fold validation4.3.5 Generating predictions on new data4.3.6 Wrapping up
5 Fundamentals of machine learning
5.1 Generalization: The goal of machine learning5.1.1 Underfitting and overfitting5.1.2 The nature of generalization in deep learning5.2 Evaluating machine-learning models5.2.1 Training, validation, and test sets5.2.2 Beating a common-sense baseline5.2.3 Things to keep in mind about model evaluation5.3 Improving model fit5.3.1 Tuning key gradient descent parameters5.3.2 Using better architecture priors5.3.3 Increasing model capacity5.4 Improving generalization5.4.1 Dataset curation5.4.2 Feature engineering5.4.3 Using early stopping5.4.4 Regularizing your model
6 The universal workflow of machine learning
6.1 Defining the task6.1.1 Framing the problem6.1.2 Collecting a dataset6.1.3 Understanding your data6.1.4 Choosing a measure of success6.2 Developing a model6.2.1 Preparing the data6.2.2 Choosing an evaluation protocol6.2.3 Beating a baseline6.2.4 Scaling up: Developing a model that overfits6.2.5 Regularizing and tuning your model6.3 Deploying your model6.3.1 Explaining your work to stakeholders and setting expectations6.3.2 Shipping an inference model6.3.3 Monitoring your model in the wild6.3.4 Maintaining your model
7 A deep dive on Keras
7.1 A spectrum of workflows7.2 Different ways to build Keras models7.2.1 The Sequential model7.2.2 The Functional API7.2.3 Subclassing the Model class7.2.4 Mixing and matching different components7.2.5 Remember: Use the right tool for the job7.3 Using built-in training and evaluation loops7.3.1 Writing your own metrics7.3.2 Using callbacks7.3.3 Writing your own callbacks7.3.4 Monitoring and visualization with TensorBoard7.4 Writing your own training and evaluation loops7.4.1 Training vs. inference7.4.2 Writing custom training step functions7.4.3 Low-level usage of metrics7.4.4 Using fit() with a custom training loop7.4.5 Handling metrics in a custom train_step()
8 Image classification
8.1 Introduction to ConvNets8.1.1 The convolution operation8.1.2 The max-pooling operation8.2 Training a ConvNet from scratch on a small dataset8.2.1 The relevance of deep learning for small-data problems8.2.2 Downloading the data8.2.3 Building your model8.2.4 Data preprocessing8.2.5 Using data augmentation8.3 Using a pretrained model8.3.1 Feature extraction with a pretrained model8.3.2 Fine-tuning a pretrained model
9 ConvNet architecture patterns
9.1 Modularity, hierarchy, and reuse9.2 Residual connections9.3 Batch normalization9.4 Depthwise separable convolutions9.5 Putting it together: A mini Xception-like model9.6 Beyond convolution: Vision Transformers
10 Interpreting what ConvNets learn
10.1 Visualizing intermediate activations10.2 Visualizing ConvNet filters10.2.1 Gradient ascent in TensorFlow10.2.2 Gradient ascent in PyTorch10.2.3 Gradient ascent in JAX10.2.4 The filter visualization loop10.3 Visualizing heatmaps of class activation10.3.1 Getting the gradient of the top class: TensorFlow version10.3.2 Getting the gradient of the top class: PyTorch version10.3.3 Getting the gradient of the top class: JAX version10.3.4 Displaying the class activation heatmap10.4 Visualizing the latent space of a ConvNet
11 Image segmentation
11.1 Computer vision tasks11.1.1 Types of image segmentation11.2 Training a segmentation model from scratch11.2.1 Downloading a segmentation dataset11.2.2 Building and training the segmentation model11.3 Using a pretrained segmentation model11.3.1 Downloading the Segment Anything Model11.3.2 How Segment Anything works11.3.3 Preparing a test image11.3.4 Prompting the model with a target point11.3.5 Prompting the model with a target box
12 Object detection
12.1 Single-stage vs. two-stage object detectors12.1.1 Two-stage R-CNN detectors12.1.2 Single-stage detectors12.2 Training a YOLO model from scratch12.2.1 Downloading the COCO dataset12.2.2 Creating a YOLO model12.2.3 Readying the COCO data for the YOLO model12.2.4 Training the YOLO model12.3 Using a pretrained RetinaNet detector
13 Timeseries forecasting
13.1 Different kinds of timeseries tasks13.2 A temperature forecasting example13.2.1 Preparing the data13.2.2 A commonsense, non-machine-learning baseline13.2.3 Let’s try a basic machine learning model13.2.4 Let’s try a 1D convolutional model13.3 Recurrent neural networks13.3.1 Understanding recurrent neural networks13.3.2 A recurrent layer in Keras13.3.3 Getting the most out of recurrent neural networks13.3.4 Using recurrent dropout to fight overfitting13.3.5 Stacking recurrent layers13.3.6 Using bidirectional RNNs13.4 Going even further
14 Text classification
14.1 A brief history of natural language processing14.2 Preparing text data14.2.1 Character and word tokenization14.2.2 Subword tokenization14.3 Sets vs. sequences14.3.1 Loading the IMDb classification dataset14.4 Set models14.4.1 Training a bag-of-words model14.4.2 Training a bigram model14.5 Sequence models14.5.1 Training a recurrent model14.5.2 Understanding word embeddings14.5.3 Using a word embedding14.5.4 Pretraining a word embedding14.5.5 Using the pretrained embedding for classification
15 Language models and the Transformer
15.1 The language model15.1.1 Training a Shakespeare language model15.1.2 Generating Shakespeare15.2 Sequence-to-sequence learning15.2.1 English-to-Spanish translation15.2.2 Sequence-to-sequence learning with RNNs15.3 The Transformer architecture15.3.1 Dot-product attention15.3.2 Transformer encoder block15.3.3 Transformer decoder block15.3.4 Sequence-to-sequence learning with a Transformer15.3.5 Embedding positional information15.4 Classification with a pretrained Transformer15.4.1 Pretraining a Transformer encoder15.4.2 Loading a pretrained Transformer15.4.3 Preprocessing IMDb movie reviews15.4.4 Fine-tuning a pretrained Transformer15.5 What makes the Transformer effective?
16 Text generation
16.1 A brief history of sequence generation16.2 Training a mini-GPT16.2.1 Building the model16.2.2 Pretraining the model16.2.3 Generative decoding16.2.4 Sampling strategies16.3 Using a pretrained LLM16.3.1 Text generation with the Gemma model16.3.2 Instruction fine-tuning16.3.3 Low-Rank Adaptation (LoRA)16.4 Going further with LLMs16.4.1 Reinforcement Learning with Human Feedback (RLHF)16.4.2 Multimodal LLMs16.4.3 Retrieval Augmented Generation (RAG)16.4.4 “Reasoning” models16.5 Where are LLMs heading next?
17 Image generation
17.1 Deep learning for image generation17.1.1 Sampling from latent spaces of images17.1.2 Variational autoencoders17.1.3 Implementing a VAE with Keras17.2 Diffusion models17.2.1 The Oxford Flowers dataset17.2.2 A U-Net denoising autoencoder17.2.3 The concepts of diffusion time and diffusion schedule17.2.4 The training process17.2.5 The generation process17.2.6 Visualizing results with a custom callback17.2.7 It’s go time!17.3 Text-to-image models17.3.1 Exploring the latent space of a text-to-image model
18 Best practices for the real world
18.1 Getting the most out of your models18.1.1 Hyperparameter optimization18.1.2 Model ensembling18.2 Scaling up model training with multiple devices18.2.1 Multi-GPU training18.2.2 Distributed training in practice18.2.3 TPU training18.3 Speeding up training and inference with lower-precision computation18.3.1 Understanding floating-point precision18.3.2 Float16 inference18.3.3 Mixed-precision training18.3.4 Using loss scaling with mixed precision18.3.5 Beyond mixed precision: float8 training18.3.6 Faster inference with quantization
19 The future of AI
19.1 The limitations of deep learning19.1.1 Deep learning models struggle to adapt to novelty19.1.2 Deep learning models are highly sensitive to phrasing and other distractors19.1.3 Deep learning models struggle to learn generalizable programs19.1.4 The risk of anthropomorphizing machine-learning models19.2 Scale isn’t all you need19.2.1 Automatons vs. intelligent agents19.2.2 Local generalization vs. extreme generalization19.2.3 The purpose of intelligence19.2.4 Climbing the spectrum of generalization19.3 How to build intelligence19.3.1 The kaleidoscope hypothesis19.3.2 The essence of intelligence: Abstraction acquisition and recombination19.3.3 The importance of setting the right target19.3.4 A new target: On-the-fly adaptation19.3.5 ARC Prize19.3.6 The test-time adaptation era19.3.7 ARC-AGI 219.4 The missing ingredients: Search and symbols19.4.1 The two poles of abstraction19.4.2 Cognition as a combination of both kinds of abstraction19.4.3 Why deep learning isn’t a complete answer to abstraction generation19.4.4 An alternative approach to AI: Program synthesis19.4.5 Blending deep learning and program synthesis19.4.6 Modular component recombination and lifelong learning19.4.7 The long-term vision
20 Conclusions
20.1 Key concepts in review20.1.1 Various approaches to artificial intelligence20.1.2 What makes deep learning special within the field of machine learning20.1.3 How to think about deep learning20.1.4 Key enabling technologies20.1.5 The universal machine learning workflow20.1.6 Key network architectures20.2 Limitations of deep learning20.3 What might lie ahead20.4 Staying up to date in a fast-moving field20.4.1 Practice on real-world problems using Kaggle20.4.2 Read about the latest developments on arXiv20.4.3 Explore the Keras ecosystem20.5 Final words

Content preview from Deep Learning with Python, Third Edition

15 Language models and the Transformer

This chapter covers

How to generate text with a deep learning model
Training a model to translate from English to Spanish
The Transformer, a powerful architecture for text modeling problems

With the basics of text preprocessing and modeling covered in the previous chapter, this chapter will tackle some more involved language problems such as machine translation. We will build up a solid intuition for the Transformer model that powers products like ChatGPT and has helped trigger a wave of investment in natural language processing (NLP).

15.1 The language model

In the previous chapter, we learned how to convert text data to numeric inputs, and we used this numeric representation to classify movie ...

Become an O’Reilly member and get unlimited access to this title plus top books and audiobooks from O’Reilly and nearly 200 top publishers, thousands of courses curated by job role, 150+ live events each month,
and much more.

Read now

Unlock full access

More than 5,000 organizations count on O’Reilly

O’Reilly covers everything we've got, with content to help us build a world-class technology community, upgrade the capabilities and competencies of our teams, and improve overall team performance as well as their engagement.

Julian F.

Head of Cybersecurity

I wanted to learn C and C++, but it didn't click for me until I picked up an O'Reilly book. When I went on the O’Reilly platform, I was astonished to find all the books there, plus live events and sandboxes so you could play around with the technology.

Addison B.

Field Engineer

I’ve been on the O’Reilly platform for more than eight years. I use a couple of learning platforms, but I'm on O'Reilly more than anybody else. When you're there, you start learning. I'm never disappointed.

Amir M.

Data Platform Tech Lead

I'm always learning. So when I got on to O'Reilly, I was like a kid in a candy store. There are playlists. There are answers. There's on-demand training. It's worth its weight in gold, in terms of what it allows me to do.

Mark W.

Embedded Software Engineer

Deep Learning with Python, Second Edition

Publisher Resources

ISBN: 9781633436589Publisher Support Publisher Website

Cloud Computing

Data Engineering

Data Science

AI & ML

Programming Languages

Software Architecture

IT/Ops

Security

Design

Business

Soft Skills

Deep Learning with Python, Third Edition

by Matthew Watson, Francois Chollet