Deep learning for dummies

Table of content
Introduction

About This Book

Foolish Assumptions

Icons Used in This Book

Beyond the Book

Where to Go from Here



Part 1: Discovering Deep Learning

Chapter 1: Introducing Deep Learning

Defining What Deep Learning Means
Starting from Artificial Intelligence
Considering the role of AI
Focusing on machine learning
Moving from machine learning to deep learning
Using Deep Learning in the Real World
Understanding the concept of learning
Performing deep learning tasks
Employing deep learning in applications
Considering the Deep Learning Programming Environment
Overcoming Deep Learning Hype
Discovering the start-up ecosystem
Knowing when not to use deep learning


Chapter 2: Introducing the Machine Learning Principles

Defining Machine Learning
Understanding how machine learning works
Understanding that it's pure math
Learning by different strategies
Training, validating, and testing data
Looking for generalization
Getting to know the limits of bias
Keeping model complexity in mind
Considering the Many Different Roads to Learning
Understanding there is no free lunch
Discovering the five main approaches
Delving into some different approaches
Awaiting the next breakthrough
Pondering the True Uses of Machine Learning
Understanding machine learning benefits
Discovering machine learning limits


Chapter 3: Getting and Using Python

Working with Python in this Book
Obtaining Your Copy of Anaconda
Getting Continuum Analytics Anaconda
Installing Anaconda on Linux
Installing Anaconda on MacOS
Installing Anaconda on Windows
Downloading the Datasets and Example Code
Using Jupyter Notebook
Defining the code repository
Getting and using datasets
Creating the Application
Understanding cells
Adding documentation cells
Using other cell types
Understanding the Use of Indentation
Adding Comments
Understanding comments
Using comments to leave yourself reminders
Using comments to keep code from executing
Getting Help with the Python Language
Working in the Cloud
Using the Kaggle datasets and kernels
Using the Google Colaboratory


Chapter 4: Leveraging a Deep Learning Framework

Presenting Frameworks
Defining the differences
Explaining the popularity of frameworks
Defining the deep learning framework
Choosing a particular framework
Working with Low-End Frameworks
Caffe2
Chainer
PyTorch
MXNet
Microsoft Cognitive Toolkit/CNTK
Understanding TensorFlow
Grasping why TensorFlow is so good
Making TensorFlow easier by using TFLearn
Using Keras as the best simplifier
Getting your copy of TensorFlow and Keras
Fixing the C++ build tools error in Windows
Accessing your new environment in Notebook


Part 2: Considering Deep Learning Basics

Chapter 5: Reviewing Matrix Math and Optimization

Revealing the Math You Really Need
Working with data
Creating and operating with a matrix
Understanding Scalar, Vector, and Matrix Operations
Creating a matrix
Performing matrix multiplication
Executing advanced matrix operations
Extending analysis to tensors
Using vectorization effectively
Interpreting Learning as Optimization
Exploring cost functions
Descending the error curve
Learning the right direction
Updating


Chapter 6: Laying Linear Regression Foundations

Combining Variables
Working through simple linear regression
Advancing to multiple linear regression
Including gradient descent
Seeing linear regression in action
Mixing Variable Types
Modeling the responses
Modeling the features
Dealing with complex relations
Switching to Probabilities
Specifying a binary response
Transforming numeric estimates into probabilities
Guessing the Right Features
Defining the outcome of incompatible features
Solving overfitting using selection and regularization
Learning One Example at a Time
Using gradient descent
Understanding how SGD is different


Chapter 7: Introducing Neural Networks

Discovering the Incredible Perceptron
Understanding perceptron functionality
Touching the nonseparability limit
Hitting Complexity with Neural Networks
Considering the neuron
Pushing data with feed-forward
Going even deeper into the rabbit hole
Using backpropagation to adjust learning
Struggling with Overfitting
Understanding the problem
Opening the black box


Chapter 8: Building a Basic Neural Network

Understanding Neural Networks
Defining the basic architecture
Documenting the essential modules
Solving a simple problem
Looking Under the Hood of Neural Networks
Choosing the right activation function
Relying on a smart optimizer
Setting a working learning rate


Chapter 9: Moving to Deep Learning

Seeing Data Everywhere
Considering the effects of structure
Understanding Moore's implications
Considering what Moore's Law changes
Discovering the Benefits of Additional Data
Defining the ramifications of data
Considering data timeliness and quality
Improving Processing Speed
Leveraging powerful hardware
Making other investments
Explaining Deep Learning Differences from Other Forms of AI
Adding more layers
Changing the activations
Adding regularization by dropout
Finding Even Smarter Solutions
Using online learning
Transferring learning
Learning end to end


Chapter 10: Explaining Convolutional Neural Networks

Beginning the CNN Tour with Character Recognition
Understanding image basics
Explaining How Convolutions Work
Understanding convolutions
Simplifying the use of pooling
Describing the LeNet architecture
Detecting Edges and Shapes from Images
Visualizing convolutions
Unveiling successful architectures
Discussing transfer learning


Chapter 11: Introducing Recurrent Neural Networks

Introducing Recurrent Networks
Modeling sequences using memory
Recognizing and translating speech
Placing the correct caption on pictures
Explaining Long Short-Term Memory
Defining memory differences
Walking through the LSTM architecture
Discovering interesting variants
Getting the necessary attention


Part 3: Interacting with Deep Learning

Chapter 12: Performing Image Classification

Using Image Classification Challenges
Delving into ImageNet and MS COCO
Learning the magic of data augmentation
Distinguishing Traffic Signs
Preparing image data
Running a classification task


Chapter 13: Learning Advanced CNNs

Distinguishing Classification Tasks
Performing localization
Classifying multiple objects
Annotating multiple objects in images
Segmenting images
Perceiving Objects in Their Surroundings
Discovering how RetinaNet works
Using the Keras-RetinaNet code
Overcoming Adversarial Attacks on Deep Learning Applications
Tricking pixels
Hacking with stickers and other artifacts


Chapter 14: Working on Language Processing

Processing Language
Defining understanding as tokenization
Putting all the documents into a bag
Memorizing Sequences that Matter
Understanding semantics by word embeddings
Using AI for Sentiment Analysis


Chapter 15: Generating Music and Visual Art

Learning to Imitate Art and Life
Transferring an artistic style
Reducing the problem to statistics
Understanding that deep learning doesn't create
Mimicking an Artist
Defining a new piece based on a single artist
Combining styles to create new art
Visualizing how neural networks dream
Using a network to compose music


Chapter 16: Building Generative Adversarial Networks

Making Networks Compete
Finding the key in the competition
Achieving more realistic results
Considering a Growing Field
Inventing realistic pictures of celebrities
Enhancing details and image translation


Chapter 17: Playing with Deep Reinforcement Learning

Playing a Game with Neural Networks
Introducing reinforcement learning
Simulating game environments
Presenting Q-learning
Explaining Alpha-Go
Determining if you're going to win
Applying self-learning at scale


Part 4: The Part of Tens

Chapter 18: Ten Applications that Require Deep Learning

Restoring Color to Black-and-White Videos and Pictures
Approximating Person Poses in Real Time
Performing Real-Time Behavior Analysis
Translating Languages
Estimating Solar Savings Potential
Beating People at Computer Games
Generating Voices
Predicting Demographics
Creating Art from Real-World Pictures
Forecasting Natural Catastrophes


Chapter 19: Ten Must-Have Deep Learning Tools

Compiling Math Expressions Using Theano
Augmenting TensorFlow Using Keras
Dynamically Computing Graphs with Chainer
Creating a MATLAB-Like Environment with Torch
Performing Tasks Dynamically with PyTorch
Accelerating Deep Learning Research Using CUDA
Supporting Business Needs with Deeplearning4j
Mining Data Using Neural Designer
Training Algorithms Using Microsoft Cognitive Toolkit (CNTK)
Exploiting Full GPU Capability Using MXNet


Chapter 20: Ten Types of Occupations that Use Deep Learning

Managing People
Improving Medicine
Developing New Devices
Providing Customer Support
Seeing Data in New Ways
Performing Analysis Faster
Creating a Better Work Environment
Researching Obscure or Detailed Information
Designing Buildings
Enhancing Safety


Index

This book makes sense of those increasingly confusing algorithms, and it creates a simple and safe environment to experiment with deep learning. It develops a sense of precisely what deep learning can do at a high level and then it provides examples of the major deep learning application types. The book includes simple example code, but there is also approachable text with real world examples, and even some hands on activities. The reader learns the topic in more than one way and from more than one perspective.