book

Deep Learning Cookbook

by Douwe Osinga

June 2018

Intermediate to advanced

252 pages

6h 9m

English

O'Reilly Media, Inc.

Read now

Unlock full access

A Brief History of Deep LearningWhy Now?What Do You Need to Know?How This Book Is StructuredConventions Used in This BookAccompanying CodeO’Reilly SafariHow to Contact UsAcknowledgments
1.1. Types of Neural Networks1.2. Acquiring Data1.3. Preprocessing Data
2.1. Determining That You Are Stuck2.2. Solving Runtime Errors2.3. Checking Intermediate Results2.4. Picking the Right Activation Function (for Your Final Layer)2.5. Regularization and Dropout2.6. Network Structure, Batch Size, and Learning Rate
3.1. Using Pretrained Word Embeddings to Find Word Similarity3.2. Word2vec Math3.3. Visualizing Word Embeddings3.4. Finding Entity Classes in Embeddings3.5. Calculating Semantic Distances Inside a Class3.6. Visualizing Country Data on a Map
4.1. Collecting the Data4.2. Training Movie Embeddings4.3. Building a Movie Recommender4.4. Predicting Simple Movie Properties
5.1. Acquiring the Text of Public Domain Books5.2. Generating Shakespeare-Like Texts5.3. Writing Code Using RNNs5.4. Controlling the Temperature of the Output5.5. Visualizing Recurrent Network Activations
6.1. Acquiring Data from Stack Exchange6.2. Exploring Data Using Pandas6.3. Using Keras to Featurize Text6.4. Building a Question/Answer Model6.5. Training a Model with Pandas6.6. Checking Similarities
7.1. Building a Simple Sentiment Classifier7.2. Inspecting a Simple Classifier7.3. Using a Convolutional Network for Sentiment Analysis7.4. Collecting Twitter Data7.5. A Simple Emoji Predictor7.6. Dropout and Multiple Windows7.7. Building a Word-Level Model7.8. Constructing Your Own Embeddings7.9. Using a Recurrent Neural Network for Classification7.10. Visualizing (Dis)Agreement7.11. Combining Models
8.1. Training a Simple Sequence-to-Sequence Model8.2. Extracting Dialogue from Texts8.3. Handling an Open Vocabulary8.4. Training a seq2seq Chatbot
9.1. Loading a Pretrained Network9.2. Preprocessing Images9.3. Running Inference on Images9.4. Using the Flickr API to Collect a Set of Labeled Images9.5. Building a Classifier That Can Tell Cats from Dogs9.6. Improving Search Results9.7. Retraining Image Recognition Networks

10.1. Acquiring Images from Wikipedia10.2. Projecting Images into an N-Dimensional Space10.3. Finding Nearest Neighbors in High-Dimensional Spaces10.4. Exploring Local Neighborhoods in Embeddings
11.1. Detecting Multiple Images Using a Pretrained Classifier11.2. Using Faster RCNN for Object Detection11.3. Running Faster RCNN over Our Own Images
12.1. Visualizing CNN Activations12.2. Octaves and Scaling12.3. Visualizing What a Neural Network Almost Sees12.4. Capturing the Style of an Image12.5. Improving the Loss Function to Increase Image Coherence12.6. Transferring the Style to a Different Image12.7. Style Interpolation
13.1. Importing Drawings from Google Quick Draw13.2. Creating an Autoencoder for Images13.3. Visualizing Autoencoder Results13.4. Sampling Images from a Correct Distribution13.5. Visualizing a Variational Autoencoder Space13.6. Conditional Variational Autoencoders
14.1. Acquiring Icons for Training14.2. Converting the Icons to a Tensor Representation14.3. Using a Variational Autoencoder to Generate Icons14.4. Using Data Augmentation to Improve the Autoencoder’s Performance14.5. Building a Generative Adversarial Network14.6. Training Generative Adversarial Networks14.7. Showing the Icons the GAN Produces14.8. Encoding Icons as Drawing Instructions14.9. Training an RNN to Draw Icons14.10. Generating Icons Using an RNN
15.1. Creating a Training Set for Music Classification15.2. Training a Music Genre Detector15.3. Visualizing Confusion15.4. Indexing Existing Music15.5. Setting Up Spotify API Access15.6. Collecting Playlists and Songs from Spotify15.7. Training a Music Recommender15.8. Recommending Songs Using a Word2vec Model
16.1. Using Scikit-Learn’s Nearest Neighbors for Embeddings16.2. Use Postgres to Store Embeddings16.3. Populating and Querying Embeddings Stored in Postgres16.4. Storing High-Dimensional Models in Postgres16.5. Writing Microservices in Python16.6. Deploying a Keras Model Using a Microservice16.7. Calling a Microservice from a Web Framework16.8. TensorFlow seq2seq models16.9. Running Deep Learning Models in the Browser16.10. Running a Keras Model Using TensorFlow Serving16.11. Using a Keras Model from iOS

Overview

Deep learning doesn’t have to be intimidating. Until recently, this machine-learning method required years of study, but with frameworks such as Keras and Tensorflow, software engineers without a background in machine learning can quickly enter the field. With the recipes in this cookbook, you’ll learn how to solve deep-learning problems for classifying and generating text, images, and music.

Each chapter consists of several recipes needed to complete a single project, such as training a music recommending system. Author Douwe Osinga also provides a chapter with half a dozen techniques to help you if you’re stuck. Examples are written in Python with code available on GitHub as a set of Python notebooks.

You’ll learn how to: