book

Deep Learning Crash Course

by Giovanni Volpe, Benjamin Midtvedt, Jesús Pineda, Henrik Klein Moberg, Harshith Bachimanchi, Joana B. Pereira, Carlo Manzo

January 2026

Intermediate to advanced

680 pages

18h 52m

English

No Starch Press

Read now

Unlock full access

Cover Page
Title Page
Copyright Page
About the Authors
About the Technical Reviewer
BRIEF CONTENTS
CONTENTS IN DETAIL
ACKNOWLEDGMENTS
INTRODUCTION
Who Should Read This Book?This Book’s ApproachSetting Up Your Learning EnvironmentObtaining the Data and Code Used in This BookWhat’s in This Book?Online ResourcesTime to Start Your Journey
1 BUILDING AND TRAINING YOUR FIRST NEURAL NETWORK
Classifying Data with a Single NeuronClassifying 1D DataClassifying 2D DataAdding a Bias as Another DimensionUsing Weight Regularization to Stabilize the TrainingUsing Different Activation FunctionsClassifying Data with a Two-Layer Neural NetworkImplementing a Dense Neural NetworkTraining with Error BackpropagationAttempting an Impossible ClassificationClassifying Data with a Three-Layer Neural NetworkProject 1A: Classifying Handwritten DigitsUsing the MNIST Digit DatasetImplementing the Neural NetworkTraining and TestingMaking ImprovementsPerforming Failure AnalysisSummarySeminal Works and Further Reading

2 CAPTURING TRENDS AND RECOGNIZING PATTERNS WITH DENSE NEURAL NETWORKS
Regressing Data with a Single NeuronRegressing 1D DataRegressing 2D DataRegressing Data with a Two-Layer Neural NetworkImplementing a Dense Neural NetworkTraining with Error BackpropagationTraining Using BatchesPlotting Predictions vs. Ground TruthTraining with Mini-BatchesTracking Training MetricsRandomizing the BatchesDividing Data into Multiple DatasetsSplitting the DataTraining and Validating the Neural NetworkProject 2A: Emulating a Physical SystemLoading the Theoretical and Geometrical–Optics Optical ForcesLoading and Preparing the Training DataImplementing the Neural NetworkImplementing Data LoadersTraining the Neural NetworkEvaluating PerformanceSummarySeminal Works and Further Reading
3 PROCESSING IMAGES WITH CONVOLUTIONAL NEURAL NETWORKS
Understanding ConvolutionsConvolving 1D DataConvolving 2D DataUsing Convolutions in a Neural NetworkImplementing Neural Networks in PyTorchDefining Convolutional LayersAdding ReLU ActivationAdding Pooling LayersUsing Upsampling LayersTransforming ImagesUsing Dense Layers to Classify ImagesProject 3A: Classifying Malaria-Infected Blood SmearsLoading the Malaria DatasetClassifying with Dense Neural NetworksClassifying with Convolutional Neural NetworksChecking the Values of the FiltersVisualizing Activations of Convolutional LayersVisualizing HeatmapsProject 3B: Localizing Microscopic ParticlesLoading the VideosManually Annotating the VideosImplementing a Convolutional Neural NetworkTraining with Annotated DataSimulating the Training DataTraining with Simulated DataProject 3C: Creating DeepDreamsLoading an ImageLoading a Pretrained Neural NetworkImplementing the DeepDreams AlgorithmUsing Multiple Layers at OnceProject 3D: Transferring the Style of ImagesLoading the Content and Style ImagesLoading a Pretrained Neural NetworkImplementing Style TransferCreating an Image in Gaudí’s StyleSummarySeminal Works and Further Reading
4 ENHANCING, GENERATING, AND ANALYZING DATA WITH AUTOENCODERS
Understanding Encoder-DecodersImplementing a Denoising Encoder-DecoderGenerating the DataCreating a Simulated DatasetDefining and Training the Encoder-DecoderDenoising the ImageChecking for Absence of Mode CollapseChecking Generalization CapabilitiesRetrainingProject 4A: Generating Images with Variational AutoencodersUnderstanding Variational AutoencodersLoading the MNIST DatasetTraining the Variational AutoencoderGenerating Images with the DecoderClustering Images with the EncoderProject 4B: Morphing Images with Wasserstein AutoencodersUnderstanding Wasserstein AutoencodersLoading the Fashion-MNIST DatasetTraining the Wasserstein AutoencoderReconstructing the Fashion-MNIST ImagesCreating New ImagesMorphing ImagesProject 4C: Detecting Anomalies in ECG DataUnderstanding Anomaly DetectionLoading the ECG DatasetDefining and Training the AutoencoderTesting with Normal and Anomalous ECGsDetecting AnomaliesSummarySeminal Works and Further Reading
5 SEGMENTING AND ANALYZING IMAGES WITH U-NETS
Introducing U-NetsUnderstanding Semantic Segmentation with U-NetsSegmenting Images of Biological TissuesLoading the Segmented Tissue ImagesCreating the Data PipelinesDefining and Training the U-NetPlotting the Training MetricsPreventing OverfittingEvaluating the Trained U-NetProject 5A: Detecting Quantum Dots in Fluorescence ImagesLoading the ImageSimulating Quantum DotsDefining and Training the U-NetEvaluating the Trained U-NetProject 5B: Counting CellsLoading the DatasetCreating a PipelineSimulating the Cell ImagesImplementing and Training the U-NetTesting the Trained U-NetSummarySeminal Works and Further Reading
6 TRAINING NEURAL NETWORKS WITH SELF-SUPERVISED LEARNING
Understanding Self-Supervised LearningSelf-Supervised Contrastive LearningSelf-Supervised Non-Contrastive LearningSelf-Supervised Geometric LearningDetermining the Position of a Particle in an ImageCreating the DatasetLearning from TranslationsLearning from Flipping the ImageImproving Performance with LodeSTARProject 6A: Localizing Mouse Stem Cells with LodeSTARUsing the Cell Tracking Challenge DatasetPreparing the Training CropCreating the Training Pipeline and Data LoaderTraining the Neural NetworkEvaluating PerformanceSummarySeminal Works and Further Reading
7 PROCESSING TIME SERIES AND LANGUAGE WITH RECURRENT NEURAL NETWORKS
Understanding Recurrent Neural NetworksUsing a Comb FilterUnderstanding a Simple Recurrent Neural NetworkPredicting Temperature with Recurrent Neural NetworksLoading the Jena Climate DatasetPreprocessing the DataImplementing a Commonsense BenchmarkDetermining the Computational DevicePredicting with a Simple Recurrent Neural NetworkStacking Multiple Recurrent LayersUsing Gated Recurrent UnitsUsing Long Short-Term Memory NetworksProject 7A: Translating with Recurrent Neural NetworksPreparing the Bilingual DatasetDefining the Sequence-to-Sequence ApplicationLoading Pretrained EmbeddingsTraining the Sequence-to-Sequence ApplicationTesting the Model PerformanceEvaluating the Model with the BLEU ScoreSummarySeminal Works and Further Reading
8 PROCESSING LANGUAGE AND CLASSIFYING IMAGES WITH ATTENTION AND TRANSFORMERS
Understanding AttentionImplementing Dot-Product AttentionVisualizing AttentionMaking the Attention Mechanism TrainableImplementing Other Attention MechanismsProject 8A: Using Attention to Improve Language TranslationIncorporating AttentionTraining and Testing the Seq2Seq Model with AttentionInterpreting the Attention MatrixProject 8B: Performing Sentiment Analysis with a TransformerBreaking Down Multi-Head AttentionUnderstanding the Transformer StructureLoading the IMDb DatasetPreprocessing the ReviewsDefining the Data LoadersBuilding an Encoder-Only TransformerTraining the ModelEvaluating the Trained ModelProject 8C: Classifying Images with a Vision TransformerUsing the CIFAR-10 DatasetData PreprocessingBuilding the ViT ModelTraining and Evaluating the ViT ModelImproving the ViT Model with CutMixUsing a Pretrained ViT ModelSummarySeminal Works and Further Reading
9 CREATING AND TRANSFORMING IMAGES WITH GENERATIVE ADVERSARIAL NETWORKS
Understanding GANsDiscriminating Between Real and Fake DataGenerating Realistic Fake DataTraining a GANGenerating Digits with a GANLoading the MNIST Dataset with PyTorchDefining the Generator and DiscriminatorTraining the GANPlotting the Intermediate ResultsPlotting the Training LossesProject 9A: Generating Digits with a Conditional GANDefining the Conditional Generator and DiscriminatorTraining the Conditional GANPlotting the Generated DigitsProject 9B: Virtually Staining a Biological TissueDownloading the Human Motor Neurons DatasetCreating a DatasetInstantiating the Generator and DiscriminatorCompiling the Conditional GANTraining the Conditional GANEvaluating the Trained Conditional GANProject 9C: Converting Between Holographic and Bright-Field Microscopy ImagesUnderstanding CycleGANsUsing the Holo2Bright DatasetInstantiating the CycleGAN Generators and DiscriminatorsTraining the CycleGANEvaluating the Trained CycleGANSummarySeminal Works and Further Reading
10 IMPLEMENTING GENERATIVE AI WITH DIFFUSION MODELS
Understanding DiffusionBreaking Down Denoising Diffusion Probabilistic ModelsModeling the Forward Diffusion ProcessDeriving the Fast Forward ProcessModeling the Reverse Diffusion ProcessGenerating Digits with a Diffusion ModelLoading the MNIST DatasetImplementing the Forward Diffusion ProcessApplying the Reverse Diffusion ProcessDefining the Positional Encoding FunctionInstantiating the Attention U-NetTraining the Diffusion ModelSampling Images from Fixed NoiseProject 10A: Generating Bespoke DigitsGuiding the Diffusion ModelDefining the Conditional Attention U-NetTraining the Conditional Diffusion ModelPlotting the Intermediate Training ResultsProject 10B: Generating Images from Text PromptsProviding a Conditional Text InputDefining the Conditional Attention U-NetTraining the Conditional Diffusion ModelPlotting the Intermediate Training ResultsTraining with the CLIP Tokenizer and CLIP Text EncoderProject 10C: Generating Super-Resolution ImagesDownloading the BioSR DatasetManaging the DatasetPreprocessing the ImagesCreating the Training and Test DatasetsAdapting the Diffusion Process for Super-ResolutionDefining the Conditional Attention U-NetTraining the Conditional Diffusion ModelSummarySeminal Works and Further Reading
11 MODELING MOLECULES AND COMPLEX SYSTEMS WITH GRAPH NEURAL NETWORKS
Understanding Graph ConvolutionsPredicting Molecular Properties with Graph ConvolutionsImplementing a Graph Convolution LayerRepresenting a Molecule as a GraphUsing the ZINC DatasetApplying a Graph Convolutional NetworkTraining the Graph Convolutional NetworkEvaluating the Graph Convolutional NetworkPredicting Molecular Properties with Message PassingImplementing a Message-Passing LayerImplementing a Message-Passing NetworkTraining and Evaluating the Message-Passing NetworkProject 11A: Simulating Complex Physical PhenomenaWorking with the SAND DatasetBuilding a Graph Network–Based SimulatorBuilding the DatasetTraining the ModelTesting the ModelSimulating the SystemProject 11B: Identifying Cell TrajectoriesExploring the Cell-Tracking DataCreating a Graph from Segmented ImagesBuilding a Training DatasetMaking MAGIKEvaluating PerformanceSummarySeminal Works and Further Reading
12 CONTINUOUSLY IMPROVING PERFORMANCE WITH ACTIVE LEARNING
Understanding Active LearningPerforming Binary ClassificationCreating a Dataset with Two Groups of Data PointsClassifying the Data Points with a Logistic RegressionImplementing the Active Learning ProcessComparing Random and Uncertainty Sampling StrategiesPerforming Multiclass ClassificationCreating a Dataset with Three Groups of Data PointsImplementing the Active Learning Process with Multiple ClassesComparing Sampling StrategiesProject 12A: Classifying MNIST Digits with Active LearningTraining a Baseline ModelImplementing Multiple Active Learning StrategiesComparing the Performance of the Active Learning StrategiesSummarySeminal Works and Further Reading
13 MASTERING DECISION-MAKING WITH DEEP REINFORCEMENT LEARNING
Understanding Reinforcement Learning and Q-LearningImplementing TetrisConstructing a Simplified TetrisPlaying Tetris with the Command LinePlaying Tetris with a Pygame Graphical InterfaceMaking an Agent Play Tetris with Q-LearningAdapting the Actions of Tetris for Q-LearningImplementing the Q-Learning AgentTraining the Q-Learning AgentTraining with a Random Tile SequenceMaking an Agent Play Tetris with Deep Q-LearningUnderstanding Deep Q-LearningImplementing the Deep Q-Learning AgentTraining the Deep Q-Learning AgentSummarySeminal Works and Further Reading
14 PREDICTING CHAOS WITH RESERVOIR COMPUTING
Introducing Reservoir ComputingDefining the Lorenz SystemNumerically Integrating the Lorenz SystemVisualizing Time EvolutionVisualizing the Lorenz AttractorDemonstrating the Butterfly EffectImplementing a Reservoir ComputerSetting Up the ReservoirPreparing the Training and Validation DataTraining the Reservoir ComputerEvaluating the Performance of the Reservoir ComputerSummarySeminal Works and Further Reading
CONCLUSION
INDEX

Overview

Deep Learning Crash Course is a fast-paced, thorough introduction that will have you building today’s most powerful AI models from scratch. No experience with deep learning required!

Designed for programmers who may be new to deep learning, this book offers practical, hands-on experience, not just an abstract understanding of theory.

You’ll start from the basics, and using PyTorch with real datasets, you’ll quickly progress from your first neural network to advanced architectures like convolutional neural networks (CNNs), transformers, diffusion models, and graph neural networks (GNNs). Each project can be run on your own hardware or in the cloud, with annotated code available on GitHub.

You’ll build and train models to:

Classify and analyze images, sequences, and time series
Generate and transform data with autoencoders, GANs (generative adversarial networks), and diffusion models
Process natural language with recurrent neural networks and transformers
Model molecules and physical systems with graph neural networks
Improve continuously through reinforcement and active learning
Predict chaotic systems with reservoir computing

Whether you’re an engineer, scientist, or professional developer, you’ll gain fluency in deep learning and the confidence to apply it to ambitious, real-world problems. With Deep Learning Crash Course, you’ll move from using AI tools to creating them.

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

Publisher Resources

ISBN: 0642572230210Errata Page

Cloud Computing

Data Engineering

Data Science

AI & ML

Programming Languages

Software Architecture

IT/Ops

Security

Design

Business

Soft Skills