Log In
Or create an account ->
Imperial Library
Home
About
News
Upload
Forum
Help
Login/SignUp
Index
Beginning Application Development with TensorFlow and Keras
Table of Contents
Beginning Application Development with TensorFlow and Keras
Why subscribe?
PacktPub.com
Contributors
About the author
About the reviewer
Preface
What This Book Covers
What You Need for This Book
Who This Book is for
Conventions
Reader Feedback
Customer Support
Downloading the Example Code
Installation
Installing Visual Studio
Installing Python 3
Installing TensorFlow
Installing Keras
Errata
Piracy
Questions
1. Introduction to Neural Networks and Deep Learning
Lesson Objectives
What are Neural Networks?
Successful Applications
Why Do Neural Networks Work So Well?
Representation Learning
Function Approximation
Limitations of Deep Learning
Inherent Bias and Ethical Considerations
Common Components and Operations of Neural Networks
Configuring a Deep Learning Environment
Software Components for Deep Learning
Python 3
TensorFlow
Keras
TensorBoard
Jupyter Notebooks, Pandas, and NumPy
Activity 1 – Verifying Software Components
Exploring a Trained Neural Network
MNIST Dataset
Training a Neural Network with TensorFlow
Training a Neural Network
Testing Network Performance with Unseen Data
Activity 2 – Exploring a Trained Neural Network
Summary
2. Model Architecture
Lesson Objectives
Choosing the Right Model Architecture
Common Architectures
Convolutional Neural Networks
Recurrent Neural Networks
Generative Adversarial Networks
Deep Reinforcement Learning
Data Normalization
Z-score
Point-Relative Normalization
Maximum and Minimum Normalization
Structuring Your Problem
Activity 3 – Exploring the Bitcoin Dataset and Preparing Data for Model
Using Keras as a TensorFlow Interface
Model Components
Activity 4 – Creating a TensorFlow Model Using Keras
From Data Preparation to Modeling
Training a Neural Network
Reshaping Time-Series Data
Making Predictions
Overfitting
Activity 5 – Assembling a Deep Learning System
Summary
3. Model Evaluation and Optimization
Lesson Objectives
Model Evaluation
Problem Categories
Loss Functions, Accuracy, and Error Rates
Different Loss Functions, Same Architecture
Using TensorBoard
Implementing Model Evaluation Metrics
Evaluating the Bitcoin Model
Overfitting
Model Predictions
Interpreting Predictions
Activity 6 – Creating an Active Training Environment
Hyperparameter Optimization
Layers and Nodes - Adding More Layers
Adding More Nodes
Layers and Nodes - Implementation
Epochs
Epochs - Implementation
Activation Functions
Linear (Identity)
Hyperbolic Tangent (Tanh)
Rectified Linear Unit
Activation Functions - Implementation
Regularization Strategies
L2 Regularization
Dropout
Regularization Strategies – Implementation
Optimization Results
Activity 7 – Optimizing a Deep Learning Model
Summary
4. Productization
Lesson Objectives
Handling New Data
Separating Data and Model
Data Component
Model Component
Dealing with New Data
Re-Training an Old Model
Training a New Model
Activity 8 – Dealing with New Data
Deploying a Model as a Web Application
Application Architecture and Technologies
Deploying and Using Cryptonic
Activity 9 – Deploying a Deep Learning Application
Summary
Index
← Prev
Back
Next →
← Prev
Back
Next →