Muutke küpsiste eelistusi

Generative Deep Learning: Teaching Machines To Paint, Write, Compose, and Play 2nd New edition [Pehme köide]

4.30/5 (258 hinnangut Goodreads-ist)
  • Formaat: Paperback / softback, 453 pages, kõrgus x laius: 233x178 mm
  • Ilmumisaeg: 12-May-2023
  • Kirjastus: O'Reilly Media
  • ISBN-10: 1098134184
  • ISBN-13: 9781098134181
Teised raamatud teemal:
  • Pehme köide
  • Hind: 75,81 €*
  • * hind on lõplik, st. muud allahindlused enam ei rakendu
  • Tavahind: 89,19 €
  • Säästad 15%
  • Raamatu kohalejõudmiseks kirjastusest kulub orienteeruvalt 2-4 nädalat
  • Kogus:
  • Lisa ostukorvi
  • Tasuta tarne
  • Tellimisaeg 2-4 nädalat
  • Lisa soovinimekirja
Generative Deep Learning: Teaching Machines To Paint, Write, Compose, and Play 2nd New editionSuurem pilt
  • Formaat: Paperback / softback, 453 pages, kõrgus x laius: 233x178 mm
  • Ilmumisaeg: 12-May-2023
  • Kirjastus: O'Reilly Media
  • ISBN-10: 1098134184
  • ISBN-13: 9781098134181
Teised raamatud teemal:
Püsilink: https://www.kriso.ee/db/9781098134181.html
Märksõnad:

Generative AI is the hottest topic in tech. This practical book teaches machine learning engineers and data scientists how to use TensorFlow and Keras to create impressive generative deep learning models from scratch, including variational autoencoders (VAEs), generative adversarial networks (GANs), Transformers, normalizing flows, energy-based models, and denoising diffusion models.

The book starts with the basics of deep learning and progresses to cutting-edge architectures. Through tips and tricks, you'll understand how to make your models learn more efficiently and become more creative.

  • Discover how VAEs can change facial expressions in photos
  • Train GANs to generate images based on your own dataset
  • Build diffusion models to produce new varieties of flowers
  • Train your own GPT for text generation
  • Learn how large language models like ChatGPT are trained
  • Explore state-of-the-art architectures such as StyleGAN2 and ViT-VQGAN
  • Compose polyphonic music using Transformers and MuseGAN
  • Understand how generative world models can solve reinforcement learning tasks
  • Dive into multimodal models such as DALL.E 2, Imagen, and Stable Diffusion

This book also explores the future of generative AI and how individuals and companies can proactively begin to leverage this remarkable new technology to create competitive advantage.

Foreword xv
Preface xvii
Part I Introduction to Generative Deep Learning
1 Generative Modeling
3(20)
What Is Generative Modeling?
4(1)
Generative Versus Discriminative Modeling
5(1)
The Rise of Generative Modeling
6(2)
Generative Modeling and AI
8(1)
Our First Generative Model
9(1)
Hello World!
9(1)
The Generative Modeling Framework
10(2)
Representation Learning
12(3)
Core Probability Theory
15(3)
Generative Model Taxonomy
18(2)
The Generative Deep Learning Codebase
20(1)
Cloning the Repository
20(1)
Using Docker
21(1)
Running on a GPU
21(1)
Summary
21(2)
2 Deep Learning
23(36)
Data for Deep Learning
24(1)
Deep Neural Networks
25(1)
What Is a Neural Network?
25(1)
Learning High-Level Features
26(1)
Tensor Flow and Keras
27(1)
Multilayer Perceptron (MLP)
28(1)
Preparing the Data
28(2)
Building the Model
30(5)
Compiling the Model
35(2)
Training the Model
37(1)
Evaluating the Model
38(2)
Convolutional Neural Network (CNN)
40(1)
Convolutional Layers
41(5)
Batch Normalization
46(3)
Dropout
49(2)
Building the CNN
51(2)
Training and Evaluating the CNN
53(1)
Summary
54(5)
Part II Methods
3 Variational Autoencoders
59(36)
Introduction
60(1)
Autoencoders
61(1)
The Fashion-MNIST Dataset
62(1)
The Autoencoder Architecture
63(1)
The Encoder
64(1)
The Decoder
65(2)
Joining the Encoder to the Decoder
67(2)
Reconstructing Images
69(1)
Visualizing the Latent Space
70(1)
Generating New Images
71(3)
Variational Autoencoders
74(1)
The Encoder
75(5)
The Loss Function
80(2)
Training the Variational Autoencoder
82(2)
Analysis of the Variational Autoencoder
84(1)
Exploring the Latent Space
85(1)
The CelebA Dataset
85(2)
Training the Variational Autoencoder
87(2)
Analysis of the Variational Autoencoder
89(1)
Generating New Faces
90(1)
Latent Space Arithmetic
91(1)
Morphing Between Faces
92(1)
Summary
93(2)
4 Generative Adversarial Networks
95(34)
Introduction
96(1)
Deep Convolutional GAN (DCGAN)
97(1)
The Bricks Dataset
98(1)
The Discriminator
99(2)
The Generator
101(3)
Training the DCGAN
104(5)
Analysis of the DCGAN
109(1)
GAN Training: Tips and Tricks
110(3)
Wasserstein GAN with Gradient Penalty (WGAN-GP)
113(1)
Wasserstein Loss
114(1)
The Lipschitz Constraint
115(1)
Enforcing the Lipschitz Constraint
116(1)
The Gradient Penalty Loss
117(2)
Training the WGAN-GP
119(2)
Analysis of the WGAN-GP
121(1)
Conditional GAN (CGAN)
122(1)
CGAN Architecture
123(1)
Training the CGAN
124(2)
Analysis of the CGAN
126(1)
Summary
127(2)
5 Autoregressive Models
129(38)
Introduction
130(1)
Long Short-Term Memory Network (LSTM)
131(1)
The Recipes Dataset
132(1)
Working with Text Data
133(1)
Tokenization
134(3)
Creating the Training Set
137(1)
The LSTM Architecture
138(1)
The Embedding Layer
138(2)
The LSTM Layer
140(2)
The LSTM Cell
142(2)
Training the LSTM
144(2)
Analysis of the LSTM
146(3)
Recurrent Neural Network (RNN) Extensions
149(1)
Stacked Recurrent Networks
149(2)
Gated Recurrent Units
151(2)
Bidirectional Cells
153(1)
Pixel CNN
153(1)
Masked Convolutional Layers
154(2)
Residual Blocks
156(2)
Training the Pixel CNN
158(1)
Analysis of the Pixel CNN
159(3)
Mixture Distributions
162(2)
Summary
164(3)
6 Normalizing Flow Models
167(22)
Introduction
168(1)
Normalizing Flows
169(1)
Change of Variables
170(2)
The Jacobian Determinant
172(1)
The Change of Variables Equation
173(1)
Real NVP
174(1)
The Two Moons Dataset
174(1)
Coupling Layers
175(6)
Training the Real NVP Model
181(3)
Analysis of the Real NVP Model
184(2)
Other Normalizing Flow Models
186(1)
GLOW
186(1)
FFJORD
187(1)
Summary
188(1)
7 Energy-Based Models
189(16)
Introduction
189(2)
Energy-Based Models
191(1)
The MNIST Dataset
192(1)
The Energy Function
193(1)
Sampling Using Langevin Dynamics
194(3)
Training with Contrastive Divergence
197(4)
Analysis of the Energy-Based Model
201(1)
Other Energy-Based Models
202(1)
Summary
203(2)
8 Diffusion Models
205(30)
Introduction
206(2)
Denoising Diffusion Models (DDM)
208(1)
The Flowers Dataset
208(1)
The Forward Diffusion Process
209(1)
The Reparameterization Trick
210(1)
Diffusion Schedules
211(3)
The Reverse Diffusion Process
214(3)
The U-Net Denoising Model
217(7)
Training the Diffusion Model
224(1)
Sampling from the Denoising Diffusion Model
225(3)
Analysis of the Diffusion Model
228(3)
Summary
231(4)
Part III Applications
9 Transformers
235(32)
Introduction
236(1)
GPT
236(1)
The Wine Reviews Dataset
237(1)
Attention
238(1)
Queries, Keys, and Values
239(2)
Multihead Attention
241(1)
Causal Masking
242(3)
The Transformer Block
245(3)
Positional Encoding
248(2)
Training GPT
250(2)
Analysis of GPT
252(3)
Other Transformers
255(1)
T5
256(3)
GPT-3 and GPT-4
259(1)
ChatGPT
260(4)
Summary
264(3)
10 Advanced GANs
267(30)
Introduction
268(1)
ProGAN
269(1)
Progressive Training
269(7)
Outputs
276(1)
StyleGAN
277(1)
The Mapping Network
278(1)
The Synthesis Network
279(1)
Outputs from StyleGAN
280(1)
StyleGAN2
281(1)
Weight Modulation and Demodulation
282(1)
Path Length Regularization
283(1)
No Progressive Growing
284(2)
Outputs from StyleGAN2
286(1)
Other Important GANs
286(1)
Self-Attention GAN (SAGAN)
286(2)
BigGAN
288(1)
VQ-GAN
289(3)
ViT VQ-GAN
292(2)
Summary
294(3)
11 Music Generation
297(34)
Introduction
298(1)
Transformers for Music Generation
299(1)
The Bach Cello Suite Dataset
300(1)
Parsing MIDI Files
300(3)
Tokenization
303(1)
Creating the Training Set
304(1)
Sine Position Encoding
305(2)
Multiple Inputs and Outputs
307(2)
Analysis of the Music-Generating Transformer
309(4)
Tokenization of Polyphonic Music
313(4)
MuseGAN
317(1)
The Bach Chorale Dataset
317(3)
The MuseGAN Generator
320(6)
The MuseGAN Critic
326(1)
Analysis of the MuseGAN
327(2)
Summary
329(2)
12 World Models
331(28)
Introduction
331(1)
Reinforcement Learning
332(2)
The CarRacing Environment
334(2)
World Model Overview
336(1)
Architecture
336(2)
Training
338(1)
Collecting Random Rollout Data
339(2)
Training the VAE
341(1)
The VAE Architecture
341(2)
Exploring the VAE
343(3)
Collecting Data to Train the MDN-RNN
346(1)
Training the MDN-RNN
346(1)
The MDN-RNN Architecture
347(1)
Sampling from the MDN-RNN
348(1)
Training the Controller
348(1)
The Controller Architecture
349(1)
CMA-ES
349(2)
Parallelizing CMA-ES
351(2)
In-Dream Training
353(3)
Summary
356(3)
13 Multimodal Models
359(32)
Introduction
360(1)
DALL.E 2
361(1)
Architecture
362(1)
The Text Encoder
362(1)
CLIP
362(5)
The Prior
367(2)
The Decoder
369(4)
Examples from DALL.E 2
373(4)
Imagen
377(1)
Architecture
377(1)
DrawBench
378(1)
Examples from Imagen
379(1)
Stable Diffusion
380(1)
Architecture
380(1)
Examples from Stable Diffusion
381(1)
Flamingo
381(1)
Architecture
382(1)
The Vision Encoder
382(1)
The Perceiver Resampler
383(2)
The Language Model
385(3)
Examples from Flamingo
388(1)
Summary
389(2)
14 Conclusion
391(26)
Timeline of Generative AI
392(2)
2014-2017: The VAE and GAN Era
394(1)
2018-2019: The Transformer Era
394(1)
2020-2022: The Big Model Era
395(1)
The Current State of Generative AI
396(1)
Large Language Models
396(4)
Text-to-Code Models
400(2)
Text-to-image Models
402(3)
Other Applications
405(2)
The Future of Generative AI
407(1)
Generative AI in Everyday Life
407(2)
Generative AI in the Workplace
409(1)
Generative AI in Education
410(1)
Generative AI Ethics and Challenges
411(2)
Final Thoughts
413(4)
Index 417
David Foster is a Founding Partner of ADSP, a consultancy delivering bespoke data science and AI solutions. He holds an MA in Mathematics from Trinity College, Cambridge and an MSc in Operational Research from the University of Warwick. Through ADSP, David leads the delivery of high-profile data science and AI projects across the public and private sectors. He has won several international machine-learning competitions, including the Innocentive Predicting Product Purchase challenge and for delivering a process to enable a pharmaceutical company in the US to optimize site selection for clinical trials. He is a member of the Machine Learning Institute Faculty and has given talks internationally on topics related to the application of cutting-edge data science and AI within industry and academia.