Muutke küpsiste eelistusi

Deep Learning with R, Second Edition 2nd edition [Pehme köide]

4.45/5 (87 hinnangut Goodreads-ist)
, ,
  • Formaat: Paperback / softback, 568 pages, kõrgus x laius x paksus: 234x186x32 mm, kaal: 1020 g
  • Ilmumisaeg: 03-Aug-2022
  • Kirjastus: Manning Publications
  • ISBN-10: 1633439844
  • ISBN-13: 9781633439849
Teised raamatud teemal:
Deep Learning with R, Second Edition 2nd editionSuurem pilt
  • Formaat: Paperback / softback, 568 pages, kõrgus x laius x paksus: 234x186x32 mm, kaal: 1020 g
  • Ilmumisaeg: 03-Aug-2022
  • Kirjastus: Manning Publications
  • ISBN-10: 1633439844
  • ISBN-13: 9781633439849
Teised raamatud teemal:
Püsilink: https://www.kriso.ee/db/9781633439849.html
Märksõnad:
Deep learning from the ground up using R and the powerful Keras library!



In   Deep Learning with R, Second Edition  you will learn:





Deep learning from first principles Image classification and image segmentation Time series forecasting Text classification and machine translation Text generation, neural style transfer, and image generation

Deep Learning with R, Second Edition  shows you how to put deep learning into action. It's based on the revised new edition of François Chollet's bestselling   Deep Learning with Python. All code and examples have been expertly translated to the R language by Tomasz Kalinowski, who maintains the Keras and Tensorflow R packages at RStudio. Novices and experienced ML practitioners will love the expert insights, practical techniques, and important theory for building neural networks. about the technology Deep learning has become essential knowledge for data scientists, researchers, and software developers. The R language APIs for Keras and TensorFlow put deep learning within reach for all R users, even if they have no experience with advanced machine learning or neural networks. This book shows you how to get started on core DL tasks like computer vision, natural language processing, and more using R. what's inside

Image classification and image segmentation Time series forecasting Text classification and machine translation Text generation, neural style transfer, and image generation

about the reader For readers with intermediate R skills. No previous experience with Keras, TensorFlow, or deep learning is required.  
Preface xii
Acknowledgments xiv
About this book xv
About The Authors xviii
1 What is deep learning?
1(25)
1.1 Artificial intelligence, machine learning, and deep learning
2(11)
Artificial intelligence
2(1)
Machine learning
3(1)
Learning rules and representations from data
4(3)
The "deep" in "deep learning"
7(3)
Understanding how deep learning works, in three figures 8m What deep learning has achieved so far
10(1)
Don't believe the short-term hype
11(1)
The promise of AI
12(1)
1.2 Before deep learning: A brief history of machine learning
13(7)
Probabilistic modeling
13(1)
Early neural networks
13(1)
Kernel methods
14(1)
Decision trees, random forests, and gradient-boosting machines
15(1)
Back to neural networks
16(1)
What makes deep learning different?
17(1)
The modern machine learning landscape
17(3)
1.3 Why deep learning? Why now?
20(6)
Hardware
20(1)
Data
21(1)
Algorithms
22(1)
A new wave of investment
22(1)
The democratization of deep learning
23(1)
Will it last?
24(2)
2 The mathematical building blocks of neural networks
26(77)
2.1 A first look at a neural network
27(4)
2.2 Data representations for neural networks
31(6)
Scalars (rank 0 tensors)
31(1)
Vectors (rank 1 tensors)
31(1)
Matrices (rank 2 tensors)
32(1)
Rank 3 and higher-rank tensors
32(1)
Key attributes
33(1)
Manipulating tensors in R
34(1)
The notion of data batches
35(1)
Real-world examples of data tensors
35(1)
Vector data
35(1)
Time-series data or sequence data
36(1)
Image data
36(1)
Video data
37(1)
2.3 The gears of neural networks: Tensor operations
37(11)
Element-wise operations
38(2)
Broadcasting
40(1)
Tensor product
41(2)
Tensor reshaping
43(1)
Geometric interpretation of tensor operations
44(3)
A geometric interpretation of deep learning
47(1)
2.4 The engine of neural networks: Gradient-based optimization
48(11)
What's a derivative?
49(1)
Derivative of a tensor operation: The gradient
50(1)
Stochastic gradient descent
51(3)
Chaining derivatives: The backpropagation algorithm
54(5)
2.5 Looking back at our first example
59(10)
Reimplementing our first example from scratch in TensorFlow
61(2)
Running one training step
63(2)
The full training loop
65(1)
Evaluating the model
66(2)
Introduction to Keras and TensorFlow
68(1)
3.1 What's TensorFlow?
69(1)
3.2 What's Keras?
69(2)
3.3 Keras and TensorFlow: A brief history
71(1)
3.4 Python and R interfaces: A brief history
71(1)
3.5 Setting up a deep learning workspace
72(2)
Installing Keras and TensorFlow
73(1)
3.6 First steps with TensorFlow
74(1)
TensorFlow tensors
74(1)
3.7 Tensor attributes
75(14)
Tensor shape and reshaping
77(1)
Tensor slicing
78(1)
Tensor broadcasting
79(1)
The tf module
80(1)
Constant tensors and variables
81(1)
Tensor operations: Doing math in TensorFlow
82(1)
A second look at the GradientTape API
83(1)
An end-to-end example: A linear classifier in pure TensorFlow
84(5)
3.8 Anatomy of a neural network: Understanding core Keras APIs
89(14)
Layers: The building blocks of deep learning
89(5)
From layers to models
94(1)
The "compile" step: Configuring the learning process
95(3)
Picking a loss function
98(1)
Understanding the fit() method
99(1)
Monitoring loss and metrics on validation data
99(2)
Inference: Using a model after training
101(2)
4 Getting started with neural networks: Classification 1 and regression
103(27)
4.1 Classifying movie reviews: A binary classification example
105(9)
The IMDB dataset
105(2)
Preparing the data
107(1)
Building your model
108(2)
Validating your approach
110(3)
Using a trained model to generate predictions on new data
113(1)
Further experiments
113(1)
Wrapping up
113(1)
4.2 Classifying newswires: A multiclass classification example
114(8)
The Reuters dataset
114(2)
Preparing the data
116(1)
Building your model
116(1)
Validating your approach
117(2)
Generating predictions on new data
119(1)
A different way to handle the labels and the loss
120(1)
The importance of having sufficiently large intermediate layers
120(1)
Further experiments
121(1)
Wrapping up
121(1)
4.3 Predicting house prices: A regression example
122(8)
The Boston housing price dataset
122(1)
Preparing the data
123(1)
Building your model
123(1)
Validatingyour approach using K-fold validation
124(4)
Generating predictions on new data
128(1)
Wrapping up
128(2)
5 Fundamentals of machine learning
130(36)
5.1 Generalization: The goal of machine learning
130(12)
Underfilling and overfitting
131(5)
The nature of generalization in deep learning
136(6)
5.2 Evaluating machine learning models
142(4)
Training, validation, and test sets
142(3)
Beating a common-sense baseline
145(1)
Things to keep in mind about model evaluation
146(1)
5.3 Improving model fit
146(6)
Tuning key gradient descent parameters
147(2)
Leveraging better architecture priors
149(1)
Increasing model capacity
150(2)
5.4 Improving generalization
152(14)
Dataset curation
152(1)
Feature engineering
153(1)
Using early stopping
154(1)
Regularizing your model
155(11)
6 The universal workflow of machine learning
166(19)
6.1 Define the task
168(6)
Frame the problem
168(1)
Collect a dataset
169(4)
Understand your data
173(1)
Choose a measure of success
173(1)
6.2 Develop a model
174(4)
Prepare the data
174(1)
Choose an evaluation protocol
175(1)
Beat a baseline
176(1)
Scale up: Develop a model that over/its
177(1)
Regularize and tune your model
177(1)
6.3 Deploy the model
178(7)
Explain your work to stakeholders and set expectations
178(1)
Ship an inference model
179(3)
Monitor your model in the wild
182(1)
Maintain your model
183(2)
7 Working with Keras: A deep dive
185(35)
7.1 A spectrum of workflows
186(1)
7.2 Different ways to build Keras models
186(15)
The Sequential model
187(2)
The Functional API
189(7)
Subclassing the Model class
196(3)
Mixing and matching different components
199(1)
Remember: Use the right tool for the job
200(1)
7.3 Using built-in training and evaluation loops
201(9)
Writing your own metrics
202(2)
Using callbacks
204(1)
Writing your own callbacks
205(3)
Monitoring and visualization with TensorBoard
208(2)
7.4 Writing your own training and evaluation loops
210(10)
Training vs. inference
210(1)
Low-level usage of metrics
211(1)
A complete training and evaluation loop
212(3)
Make it fast with tfJunction()
215(1)
Leveraging fit() with a custom training loop
216(4)
8 Introduction to deep learning for computer vision
220(38)
8.1 Introduction to convnets
221(9)
The convolution operation
223(5)
The max-pooling operation
228(2)
8.2 Training a convnet from scratch on a small dataset
230(15)
The relevance of deep learning for small data problems
230(1)
Downloading the data
231(3)
Building the model
234(1)
Data preprocessing
235(6)
Using data augmentation
241(4)
8.3 Leveraging a pretrained model
245(13)
Feature extraction with a pretrained model
246(8)
Fine-tuning a pretrained model
254(4)
9 Advanced deep learning for computer vision
258(43)
9.1 Three essential computer vision tasks
259(1)
9.2 An image segmentation example
260(9)
9.3 Modern convnet architecture patterns
269(13)
Modularity, hierarchy, and reuse
269(3)
Residual connections
272(3)
Batch normalization
275(3)
Depthwise separable convolutions
278(2)
Putting it together: A mini Xception-like model
280(2)
9.4 Interpreting what convnets learn
282(19)
Visualizing intermediate activations
283(6)
Visualizing convnet filters
289(5)
Visualizing heatmaps of class activation
294(7)
10 Deep learning for time series
301(33)
10.1 Different kinds of time-series tasks
301(1)
10.2 A temperature-forecasting example
302(15)
Preparing the data
306(4)
A common-sense, non--machine learning baseline
310(1)
Let's try a basic machine learning model
311(3)
Let's try a ID convolutional model
314(2)
A first recurrent baseline
316(1)
10.3 Understanding recurrent neural networks
317(7)
A recurrent layer in Keras
320(4)
10.4 Advanced use of recurrent neural networks
324(10)
Using recurrent dropout to fight overfitting
324(3)
Stacking recurrent layers
327(2)
Using bidirectional RNNs
329(3)
Going even further
332(2)
11 Deep learning for text
334(65)
11.1 Natural language processing: The bird's-eye view
334(2)
11.2 Preparing text data
336(8)
Text standardization
337(1)
Text splitting (tokenization)
338(1)
Vocabulary indexing
339(1)
Using layer_text_vectorization
340(4)
11.3 Two approaches for representing groups of words: Sets and sequences
344(22)
Preparing the IMDB movie reviews data
345(2)
Processing words as a set: The bag-of-words approach
347(8)
Processing words as a sequence: The sequence model approach
355(11)
11.4 The Transformer architecture
366(16)
Understanding self-attention
366(5)
Multi-head attention
371(1)
The Transformer encoder
372(9)
When to use sequence models over bag-of-words models
381(1)
11.5 Beyond text classification: Sequence-to-sequence learning
382(17)
A machine translation example
383(4)
Sequence-to-sequence learning with RNNs
387(5)
Sequence-to-sequence learning with Transformer
392(7)
12 Generative deep learning
399(55)
12.1 Text generation
401(13)
A brief history of generative deep learning for sequence generation
401(1)
How do you generate sequence data?
402(1)
The importance of the sampling strategy
402(2)
Implementing text generation with Keras
404(4)
A text-generation callback with variable-temperature sampling
408(5)
Wrapping up
413(1)
12.2 DeepDream
414(8)
Implementing DeepDream in Keras
415(6)
Wrapping up
421(1)
12.3 Neural style transfer
422(10)
The content loss
423(1)
The style loss
424(1)
Neural style transfer in Keras
424(7)
Wrapping up
431(1)
12.4 Generating images with variational autoencoders
432(10)
Sampling from latent spaces of images
432(1)
Concept vectors for image editing
433(1)
Variational autoencoders
434(2)
Implementing a VAE with Keras
436(6)
Wrapping up
442(1)
12.5 Introduction to generative adversarial networks
442(12)
A schematic GAN implementation
443(1)
A bag of tricks
444(1)
Getting our hands on the CelebA dataset
445(2)
The discriminator
447(1)
The generator
447(1)
The adversarial network
448(4)
Wrapping up
452(2)
13 Best practices for the real world
454(19)
13.1 Getting the most out of your models
455(9)
Hyperparameter optimization
455(7)
Model ensembling
462(2)
13.2 Scaling-up model training
464(9)
Speeding up training on GPU with mixed precision
465(2)
Multi-GPU training
467(4)
TPU training
471(2)
14 Conclusions
473(38)
14.1 Key concepts in review
474(10)
Various approaches to AI
474(1)
What makes deep learning special within the field of machine learning
474(1)
How to think about deep learning
475(1)
Key enabling technologies
476(1)
The universal machine learning workflow
477(1)
Key network architectures
478(4)
The space of possibilities
482(2)
14.2 The limitations of deep learning
484(8)
The risk of anthropomorphizing machine learning models
485(2)
Automatons vs. intelligent agents
487(1)
Local generalization vs. extreme generalization
488(2)
The purpose of intelligence
490(1)
Climbing the. spectrum of generalization
491(1)
14.3 Setting the course toward greater generality in AI
492(3)
On the importance of setting the right objective: The shortcut rule
492(2)
A new target
494(1)
14.4 Implementing intelligence: The missing ingredients
495(6)
Intelligence as sensitivity to abstract analogies
496(1)
The two poles of abstraction
497(3)
The two poles of abstraction
500(1)
The missing half of the picture
500(1)
14.5 The future of deep learning
501(6)
Models as programs
502(1)
Machine learning vs. program synthesis
503(1)
Blending together deep learning and program synthesis
503(2)
Lifelong learning and modular subroutine reuse
505(1)
The long-term vision
506(1)
14.6 Staying up-to-date in a fast-moving field
507(2)
Practice on real-world problems using Kaggle
508(1)
Read about the latest developments on arXiv
508(1)
Explore the Keras ecosystem
508(1)
14.7 Final words
509(2)
Appendix Python primer for R users 511(24)
Index 535
François Chollet  is a software engineer at Google and creator of Keras.  Tomasz Kalinowski  is a software engineer at RStudio and maintainer of the Keras and Tensorflow R packages.  J.J. Allaire  is the founder of RStudio, and the author of the first edition of this book.