Book Image

Python Deep Learning

By : Valentino Zocca, Gianmario Spacagna, Daniel Slater, Peter Roelants
Book Image

Python Deep Learning

By: Valentino Zocca, Gianmario Spacagna, Daniel Slater, Peter Roelants

Overview of this book

With an increasing interest in AI around the world, deep learning has attracted a great deal of public attention. Every day, deep learning algorithms are used broadly across different industries. The book will give you all the practical information available on the subject, including the best practices, using real-world use cases. You will learn to recognize and extract information to increase predictive accuracy and optimize results. Starting with a quick recap of important machine learning concepts, the book will delve straight into deep learning principles using Sci-kit learn. Moving ahead, you will learn to use the latest open source libraries such as Theano, Keras, Google's TensorFlow, and H20. Use this guide to uncover the difficulties of pattern recognition, scaling data with greater accuracy and discussing deep learning algorithms and techniques. Whether you want to dive deeper into Deep Learning, or want to investigate how to get more out of this powerful technology, you’ll find everything inside.

Related Content you might be interested in

Current Title:

Small book image
Python Deep Learning
Valentino Zocca | Gianmario Spacagna | Daniel Slater | Peter Roelants
Apr, 2017 | 406 pages
No titles found
Table of Contents (18 chapters)
Python Deep Learning
Python Deep Learning
Credits
Credits
About the Authors
About the Authors
About the Reviewer
About the Reviewer
www.PacktPub.com
www.PacktPub.com
Customer Feedback
Customer Feedback
Preface
Preface
Free Chapter
1
Machine Learning – An Introduction
Machine Learning – An Introduction
What is machine learning?
Different machine learning approaches
Summary
2
Neural Networks
Neural Networks
Why neural networks?
Fundamentals
Summary
3
Deep Learning Fundamentals
Deep Learning Fundamentals
What is deep learning?
Deep learning applications
GPU versus CPU
Popular open source libraries – an introduction
Summary
4
Unsupervised Feature Learning
Unsupervised Feature Learning
Autoencoders
Restricted Boltzmann machines
Summary
5
Image Recognition
Image Recognition
Similarities between artificial and biological models
Intuition and justification
Convolutional layers
Pooling layers
Dropout
Convolutional layers in deep learning
Convolutional layers in Theano
A convolutional layer example with Keras to recognize digits
A convolutional layer example with Keras for cifar10
Pre-training
Summary
6
Recurrent Neural Networks and Language Models
Recurrent Neural Networks and Language Models
Recurrent neural networks
Language modeling
Speech recognition
Summary
Bibliography
7
Deep Learning for Board Games
Deep Learning for Board Games
Early game playing AI
Using the min-max algorithm to value game states
Implementing a Python Tic-Tac-Toe game
Learning a value function
Training AI to master Go
Upper confidence bounds applied to trees
Deep learning in Monte Carlo Tree Search
Quick recap on reinforcement learning
Policy gradients for learning policy functions
Policy gradients in AlphaGo
Summary
8
Deep Learning for Computer Games
Deep Learning for Computer Games
A supervised learning approach to games
Applying genetic algorithms to playing games
Q-Learning
Q-learning in action
Dynamic games
Atari Breakout
Actor-critic methods
Asynchronous methods
Model-based approaches
Summary
9
Anomaly Detection
Anomaly Detection
What is anomaly and outlier detection?
Real-world applications of anomaly detection
Popular shallow machine learning techniques
Anomaly detection using deep auto-encoders
H2O
Examples
Summary
10
Building a Production-Ready Intrusion Detection System
Building a Production-Ready Intrusion Detection System
What is a data product?
Training
Testing
Model validation
Hyper-parameters tuning
End-to-end evaluation
Deployment
Summary
Index
Index

Policy gradients in AlphaGo

For AlphaGo using policy gradients, the network was set up to play games against itself. It did so with a reward of 0 for every time step until the final one where the game is either won or lost, giving a reward of 1 or -1. This final reward is then applied to every time step in the network, and the network is trained using policy gradients in the same way as our Tic-tac-toe example. To prevent overfitting, games were played against a randomly selected previous version of the network. If the network constantly plays against itself, the risk is it could end up with some very niche strategies, which would not work against varied opponents, a local minima of sorts.

Building the initial supervised learning network that predicted the most likely moves by human players allowed AlphaGo to massively reduce the breadth of the search it needs to perform in MCTS. This allowed them to get much more accurate evaluation per rollout. The problem is that running a large many-layered...

Previous Section
End of Section 11
Next Section