Deep Learning By Example

Deep Learning By Example

by Md. Rezaul Karim, Ahmed Menshawy
Released February 2018
Publisher(s): Packt Publishing
ISBN: 9781788399906

Read it now on the O’Reilly learning platform with a 10-day free trial.

O’Reilly members get unlimited access to books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.

Start your free trial

Book description

Numerical computing, data processing, and enough about deep learning to get you up and running

About This Book

  • Get your first experience with deep learning with this easy-to-follow guide

  • Implement neural networks with the easiest, developer-friendly tools and techniques in the market.

Who This Book Is For

This book is dedicated to developers, data analysts, or deep learning enthusiasts who do not have much background with complex numerical computations but want to know what is deep learning. The book majorly appeals to beginners who are looking for a quick guide to gain some hands-on experience with deep learning. Some experience with Python would be great.

What You Will Learn

  • Learn about Data Science, its challenges and how to tackle them.

  • Learn the basics of Data Science and modern best practices with a Titanic Example.

  • Get familiarized with one of the most powerful platforms for Deep Learning(DL), TensorFlow 1.x.

  • Basic of Deep Learning and modern best practices with a digit classification problem of MNIST.

  • Dive into imaging problems by looking at early lung cancer detection and emotion recognition using CNN.

  • Apply deep learning to other domains like Language Modeling, ChatBots and Machine Translation using the one of the powerful architectures of DL, RNN.

In Detail

Deep Learning has made some huge and significant contributions and it’s one of the mostly adopted techniques in order to drive insights from your data nowadays. Google developed one of the most used libraries (aka. TensorFlow) to use in order to build fast, robust against an error-prone and scale deep learning algorithms that can run on both CPU and GPU.

This book is a starting point for those who are keen on knowing about deep learning and implementing it, but do not have extensive background in machine learning. We will start with introducing you with Data science for performing data analysis, machine learning, and eventually deep learning. Then, you will explore algorithms and various techniques that lead into efficient data processing. You will learn to clean, mine, and analyze data. Once you are comfortable with some analysis, you will then move to creating machine learning models that will eventually lead you to neural networks. You will get familiar with basics of deep learning and explore various tools that enable deep learning in a powerful yet user friendly manner. While all of this is being taught, spread across the book, we will be using intuitive examples like Titanic survivor prediction, Housing price predictor, etc. teaching implementations of each of the concept. With a very low starting point, this book will enable a regular developer to get hands on experience with deep learning.

By the end of this book, you will learn all the essentials needed to explore and understand what is deep learning and will perform deep learning tasks first hand.

Table of contents

  1. Title Page
  2. Copyright and Credits
    1. Deep Learning By Example
  3. Packt Upsell
    1. Why subscribe?
    2. PacktPub.com
  4. Contributors
    1. About the author
    2. About the reviewers
    3. Packt is searching for authors like you
  5. Preface
    1. Who this book is for
    2. What this book covers
    3. To get the most out of this book
      1. Download the example code files
      2. Download the color images
      3. Conventions used
    4. Get in touch
      1. Reviews
  6. Data Science - A Birds' Eye View
    1. Understanding data science by an example
    2. Design procedure of data science algorithms
      1. Data pre-processing
        1. Data cleaning
        2. Data pre-processing
      2. Feature selection
      3. Model selection
      4. Learning process
      5. Evaluating your model
    3. Getting to learn
      1. Challenges of learning
        1. Feature extraction – feature engineering
        2. Noise
        3. Overfitting
        4. Selection of a machine learning algorithm
        5. Prior knowledge
        6. Missing values
    4. Implementing the fish recognition/detection model
      1. Knowledge base/dataset
      2. Data analysis pre-processing
      3. Model building
        1. Model training and testing
        2. Fish recognition – all together
    5. Different learning types
      1. Supervised learning
      2. Unsupervised learning
      3. Semi-supervised learning
      4. Reinforcement learning
    6. Data size and industry needs
    7. Summary
  7. Data Modeling in Action - The Titanic Example
    1. Linear models for regression
      1. Motivation
      2. Advertising – a financial example
        1. Dependencies
        2. Importing data with pandas
        3. Understanding the advertising data
        4. Data analysis and visualization
        5. Simple regression model
          1. Learning model coefficients
          2. Interpreting model coefficients
          3. Using the model for prediction
    2. Linear models for classification
      1. Classification and logistic regression
    3. Titanic example – model building and training
      1. Data handling and visualization
      2. Data analysis – supervised machine learning
    4. Different types of errors
    5. Apparent (training set) error
    6. Generalization/true error
    7. Summary
  8. Feature Engineering and Model Complexity – The Titanic Example Revisited
    1. Feature engineering
      1. Types of feature engineering
        1. Feature selection
        2. Dimensionality reduction
        3. Feature construction
      2. Titanic example revisited
        1. Missing values
          1. Removing any sample with missing values in it
          2. Missing value inputting
          3. Assigning an average value
          4. Using a regression or another simple model to predict the values of missing variables
        2. Feature transformations
          1. Dummy features
          2. Factorizing
          3. Scaling
          4. Binning
        3. Derived features
          1. Name
          2. Cabin
          3. Ticket
        4. Interaction features
    2. The curse of dimensionality
      1. Avoiding the curse of dimensionality
    3. Titanic example revisited – all together
    4. Bias-variance decomposition
    5. Learning visibility
      1. Breaking the rule of thumb
    6. Summary
  9. Get Up and Running with TensorFlow
    1. TensorFlow installation
      1. TensorFlow GPU installation for Ubuntu 16.04
        1. Installing NVIDIA drivers and CUDA 8
        2. Installing TensorFlow
      2. TensorFlow CPU installation for Ubuntu 16.04
      3. TensorFlow CPU installation for macOS X
      4. TensorFlow GPU/CPU installation for Windows
    2. The TensorFlow environment
    3. Computational graphs
    4. TensorFlow data types, variables, and placeholders
      1. Variables
      2. Placeholders
      3. Mathematical operations
    5. Getting output from TensorFlow
    6. TensorBoard – visualizing learning
    7. Summary
  10. TensorFlow in Action - Some Basic Examples
    1. Capacity of a single neuron
      1. Biological motivation and connections
    2. Activation functions
      1. Sigmoid
      2. Tanh
      3. ReLU
    3. Feed-forward neural network
    4. The need for multilayer networks
      1. Training our MLP – the backpropagation algorithm
      2. Step 1 – forward propagation
      3. Step 2 – backpropagation and weight updation
    5. TensorFlow terminologies – recap
      1. Defining multidimensional arrays using TensorFlow
      2. Why tensors?
      3. Variables
      4. Placeholders
      5. Operations
    6. Linear regression model – building and training
      1. Linear regression with TensorFlow
    7. Logistic regression model – building and training
      1. Utilizing logistic regression in TensorFlow
        1. Why use placeholders?
        2. Set model weights and bias
        3. Logistic regression model
        4. Training
        5. Cost function
    8. Summary
  11. Deep Feed-forward Neural Networks - Implementing Digit Classification
    1. Hidden units and architecture design
    2. MNIST dataset analysis
      1. The MNIST data
    3. Digit classification – model building and training
      1. Data analysis
      2. Building the model
      3. Model training
    4. Summary
  12. Introduction to Convolutional Neural Networks
    1. The convolution operation
    2. Motivation
      1. Applications of CNNs
    3. Different layers of CNNs
      1. Input layer
      2. Convolution step
      3. Introducing non-linearity
      4. The pooling step
      5. Fully connected layer
        1. Logits layer
    4. CNN basic example – MNIST digit classification
      1. Building the model
        1. Cost function
        2. Performance measures
      2. Model training
    5. Summary
  13. Object Detection – CIFAR-10 Example
    1. Object detection
    2. CIFAR-10 – modeling, building, and training
      1. Used packages
      2. Loading the CIFAR-10 dataset
      3. Data analysis and preprocessing
      4. Building the network
      5. Model training
      6. Testing the model
    3. Summary
  14. Object Detection – Transfer Learning with CNNs
    1. Transfer learning
      1. The intuition behind TL
      2. Differences between traditional machine learning and TL
    2. CIFAR-10 object detection – revisited
      1. Solution outline
      2. Loading and exploring CIFAR-10
      3. Inception model transfer values
      4. Analysis of transfer values
      5. Model building and training
    3. Summary
  15. Recurrent-Type Neural Networks - Language Modeling
    1. The intuition behind RNNs
      1. Recurrent neural networks architectures
      2. Examples of RNNs
        1. Character-level language models
          1. Language model using Shakespeare data
      3. The vanishing gradient problem
      4. The problem of long-term dependencies
    2. LSTM networks
      1. Why does LSTM work?
    3. Implementation of the language model
      1. Mini-batch generation for training
      2. Building the model
        1. Stacked LSTMs
        2. Model architecture
        3. Inputs
        4. Building an LSTM cell
        5. RNN output
        6. Training loss
        7. Optimizer
        8. Building the network
        9. Model hyperparameters
      3. Training the model
        1. Saving checkpoints
        2. Generating text
    4. Summary
  16. Representation Learning - Implementing Word Embeddings
    1. Introduction to representation learning
    2. Word2Vec
      1. Building Word2Vec model
    3. A practical example of the skip-gram architecture
    4. Skip-gram Word2Vec implementation
      1. Data analysis and pre-processing
      2. Building the model
      3. Training
    5. Summary
  17. Neural Sentiment Analysis
    1. General sentiment analysis architecture
      1. RNNs – sentiment analysis context
      2. Exploding and vanishing gradients - recap
    2. Sentiment analysis – model implementation
      1. Keras
      2. Data analysis and preprocessing
      3. Building the model
      4. Model training and results analysis
    3. Summary
  18. Autoencoders – Feature Extraction and Denoising
    1. Introduction to autoencoders
    2. Examples of autoencoders
    3. Autoencoder architectures
    4. Compressing the MNIST dataset
      1. The MNIST dataset
      2. Building the model
      3. Model training
    5. Convolutional autoencoder
      1. Dataset
      2. Building the model
      3. Model training
    6. Denoising autoencoders
      1. Building the model
      2. Model training
    7. Applications of autoencoders
      1. Image colorization
      2. More applications
    8. Summary
  19. Generative Adversarial Networks
    1. An intuitive introduction
    2. Simple implementation of GANs
      1. Model inputs
      2. Variable scope
      3. Leaky ReLU
      4. Generator
      5. Discriminator
      6. Building the GAN network
        1. Model hyperparameters
        2. Defining the generator and discriminator
        3. Discriminator and generator losses
        4. Optimizers
      7. Model training
        1. Generator samples from training
      8. Sampling from the generator
    3. Summary
  20. Face Generation and Handling Missing Labels
    1. Face generation
      1. Getting the data
      2. Exploring the Data
      3. Building the model
        1. Model inputs
        2. Discriminator
        3. Generator
        4. Model losses
        5. Model optimizer
        6. Training the model
    2. Semi-supervised learning with Generative Adversarial Networks (GANs)
      1. Intuition
      2. Data analysis and preprocessing
      3. Building the model
        1. Model inputs
        2. Generator
        3. Discriminator
        4. Model losses
        5. Model optimizer
      4. Model training
    3. Summary
  21. Implementing Fish Recognition
    1. Code for fish recognition
  22. Other Books You May Enjoy
    1. Leave a review - let other readers know what you think

Product information

  • Title: Deep Learning By Example
  • Author(s): Md. Rezaul Karim, Ahmed Menshawy
  • Release date: February 2018
  • Publisher(s): Packt Publishing
  • ISBN: 9781788399906