Hands-On Machine Learning with TensorFlow.js: A guide to building ML applications integrated with web technology using the TensorFlow.js library

Hands-On Machine Learning with TensorFlow.js

A guide to building ML applications integrated with

web technology using the TensorFlow.js library

Kai Sasaki

BIRMINGHAM - MUMBAI

Hands-On Machine Learning with TensorFlow.js

Copyright © 2019 Packt Publishing

All rights reserved. No part of this book may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, without the prior written permission of the publisher, except in the case of brief quotations embedded in critical articles or reviews.

Every effort has been made in the preparation of this book to ensure the accuracy of the information presented. However, the information contained in this book is sold without warranty, either express or implied. Neither the author, nor Packt Publishing or its dealers and distributors, will be held liable for any damages caused or alleged to have been caused directly or indirectly by this book.

Packt Publishing has endeavored to provide trademark information about all of the companies and products mentioned in this book by the appropriate use of capitals. However, Packt Publishing cannot guarantee the accuracy of this information.

Commissioning Editor: Sunith Shetty

Acquisition Editor: Devika Battike

Content Development Editor: Athikho Sapuni Rishana

Senior Editor: Sofi Rogers

Technical Editor: Joseph Sunil

Copy Editor: Safis Editing

Project Coordinator: Aishwarya Mohan

Proofreader: Safis Editing

Indexer: Priyanka Dhadke

Production Designer: Aparna Bhagat

First published: November 2019

Production reference: 1261119

Published by Packt Publishing Ltd.

Livery Place

35 Livery Street

Birmingham

B3 2PB, UK.

ISBN 978-1-83882-173-9

www.packt.com

Packt.com

Subscribe to our online digital library for full access to over 7,000 books and videos, as well as industry-leading tools to help you plan your personal development and advance your career. For more information, please visit our website.

Why subscribe?

Spend less time learning and more time coding with practical eBooks and videos from over 4,000 industry professionals

Improve your learning with skill plans built especially for you

Get a free eBook or video every month

Fully searchable for easy access to vital information

Copy and paste, print, and bookmark content

Did you know that Packt offers eBook versions of every book published, with PDF and ePub files available? You can upgrade to the eBook version at www.packt.com and, as a print book customer, you are entitled to a discount on the eBook copy. Get in touch with us at [email protected] for more details.

At www.packt.com, you can also read a collection of free technical articles, sign up for a range of free newsletters, and receive exclusive discounts and offers on Packt books and eBooks. 

Contributors

About the author

Kai Sasaki works as a software engineer at Treasure Data. He engages in developing large-scale distributed systems to make data valuable. His passion for creating artificial intelligence by processing large-scale data led him to the field of machine learning. He is one of the initial contributors to TensorFlow.js and keeps working to add new operators that are required for new types of machine learning models. Because of his work, he received the Google Open Source Peer Bonus in 2018.

About the reviewers

Edoh Kodjo graduated twice as an engineer in networking and in computer science, and is a software developer/data scientist. He started developing an interest in machine learning and artificial intelligence soon after graduating. Since then, his interest in the field has kept on growing. He currently works as a software engineer on a project for computer vision. As a frontend and JavaScript fanatic, he is a machine learning for the web advocate. Here is his motto—Life is an endless road, coding is an endless loop for a while!, which sums up how much he loves coding.

Nick Bourdakos is a developer advocate at IBM in NYC and the creator of Cloud Annotations, a collaborative open source image annotation tool for training computer vision models. His expertise is in machine learning, mainly deep learning applied to computer vision problems. He is passionate when it comes to teaching people about machine learning and has a course on YouTube regarding TensorFlow.js.

Packt is searching for authors like you

If you're interested in becoming an author for Packt, please visit authors.packtpub.com and apply today. We have worked with thousands of developers and tech professionals, just like you, to help them share their insight with the global tech community. You can make a general application, apply for a specific hot topic that we are recruiting an author for, or submit your own idea.

Table of Contents

Title Page

Copyright and Credits

Hands-On Machine Learning with TensorFlow.js

About Packt

Why subscribe?

Contributors

About the author

About the reviewers

Packt is searching for authors like you

Preface

Who this book is for

What this book covers

To get the most out of this book

Download the example code files

Download the color images

Conventions used

Get in touch

Bewertungen

Section 1: The Rationale of Machine Learning and the Usage of TensorFlow.js

Machine Learning for the Web

Technical requirements

Why machine learning on the web?

Operation graphs

Visualizing an operation graph

Automatic differentiation

What is TensorFlow.js?

Seamless integration with web technologies

Hardware acceleration using web browser APIs

Compatibility with TensorFlow

Data privacy

Installing TensorFlow.js

tfjs-converter

The low-level API

Tensors

Operations

Memory

Eager execution

The Layers API

Sequential model

Functional model

Model summary

Custom layers

Summary

Questions

Further reading

Importing Pretrained Models into TensorFlow.js

Technical requirements

The portable model format

Protocol buffers

GraphDef

NodeDef

Exporting a model from TensorFlow

TensorFlow's SavedModel

The SavedModel format

Simple save

The SavedModelBuilder API

The Keras HDF5 model

Converting models using tfjs-converter

Converting a TensorFlow SavedModel

The Keras HDF5 model

The TensorFlow Hub module

Loading the model into TensorFlow.js

Supported operations

Summary

Questions

Further reading

TensorFlow.js Ecosystem

Technical requirements

Why high-level libraries?

Using existing models

MobileNet in tfjs-models

Supported models

Image classification application

Example applications in the community

Loading the data from various kinds of storage

Data sources

Webcam

Pose detection with ML5.js

Supported models

PoseNet in ML5.js

Drawing cats with Magenta.js

Sketch drawing

XOR classification with machinelearn.js

Random forest classifier

Summary

Exercises

Further reading

Section 2: Real-World Applications of TensorFlow.js

Polynomial Regression

Technical requirements

What is polynomial regression?

Supervised learning

The simplest linear model

General polynomial model

The loss function

Optimizer for machine learning

Optimizers in TensorFlow.js

Two-dimensional curve fitting

Preparing the dataset

Applying the 2-degree polynomial model

Loss function by mean squared error

Looking into the optimization process

Fitting the curve

Summary

Questions

Further reading

Classification with Logistic Regression

Technical requirements

Background of binary classification

What is logistic regression?

The behavior of the probabilistic generative model

Optimization process

Classifying two-dimensional clusters

Preparing the dataset

Logistic regression model with the Core API

Optimizing with the cross-entropy loss function

Implementing a logistic regression model with the Layers API

Implementing a logistic regression model with machinelearn.js

Summary

Questions

Further reading

Unsupervised Learning

Technical requirements

What is unsupervised learning?

Learning how K-means works

Centroid

Algorithm

Evaluation

Generalizing K-means with the EM algorithm

The algorithm

Relationship with K-means

Clustering two groups in a 2D space

The three clusters

K-means implementation

Summary

Exercise

Further reading

Sequential Data Analysis

Technical requirements

What is Fourier transformation?

Constituent frequencies as a feature

Discrete Fourier transform

Fast Fourier transform

Cosine curve decomposition

Complex number type

The cosine curve

Fourier transformation for cosine curves

Fourier transformation for composite curves

Inversed Fourier transform

Summary

Exercise

Further reading

Dimensionality Reduction

Technical requirements

Why dimensionality reduction?

Curse of dimensionality

Understanding principal component analysis

Variance maximization

Projecting 3D points into a 2D space with PCA

Three-dimensional clusters

Principal component calculation

The variance of projected datasets

Word embedding

What is word embedding?

Loading the IMDb dataset

Embedding the model

Visualization of embeddings

Summary

Exercise

Further reading

Solving the Markov Decision Process

Technical requirements

Reinforcement learning

MDP

Discounted total reward

State-value function

Bellman equation

Q-learning

Solving the four-states environment

Designing the environment

The Q-learning process

Summary

Exercise

Further reading

Section 3: Productionizing Machine Learning Applications with TensorFlow.js

Deploying Machine Learning Applications

Technical requirements

The ecosystem around the JavaScript platform

JavaScript in modern web browsers

Node.js

Node package manager

Benefits of TypeScript in ML applications

Module bundler

Parcel

Webpack

Deploying modules with GitHub Pages

Summary

Questions

Further reading

Tuning Applications to Achieve High Performance

Technical requirements

The backend API of TensorFlow.js

Operations that use the CPU backend

Low overhead

Memory management

Implementing higher parallelism using the WebGL backend

Avoid blocking the main thread

Prefer tf.tidy to be free from memory management

Floating number precision

Save the overhead of shader compilation

Using TensorFlow with the Node.js backend

Tensor management

Tensor construction

Tensors as variables

Revisiting tensor destruction

Asynchronous data access

Profiling

Chrome profiler

Model visualization

Summary

Questions

Further reading

Future Work Around TensorFlow.js

Technical requirements

Experimental backend implementations

WebGPU – a new standard for accelerated graphics and computations

WebAssembly – where the web meets the hardware instruction set

React Native – moving forward to mobile-native applications

Creating a React Native app

Installing the dependencies required for TensorFlow.js

Writing an application

Running the application

Electron – cross-platform desktop environment

AutoML edge helper

Summary

Questions

Further Reading

Other Books You May Enjoy

Leave a review - let other readers know what you think

Preface

TensorFlow.js is a framework that enables you to create performant machine learning (ML) applications that run smoothly in a web browser. With this book, you will learn how to use TensorFlow.js to implement various ML models through an example-based approach.

In this book, you'll understand how ML models can be built on the web. Moving on, you will get to grips with using the TensorFlow.js ecosystem to develop applications more efficiently. The book will then guide you through implementing ML techniques and algorithms such as regression, clustering, fast Fourier transform (FFT), and dimensionality reduction. You will later use the Bellman equation to solve Markov decision process (MDP) problems and understand how it is related to reinforcement learning. Finally, you will explore techniques for deploying ML-based web applications and training models with TensorFlow Core. Throughout this ML book, you'll discover useful tips and tricks that will build on your knowledge.

By the end of this book, you will be equipped with the skills you need to create your own web-based ML applications and fine-tune models to achieve high performance.

Who this book is for

This book is for web developers who want to learn how to integrate ML techniques with web-based applications from scratch. This book will also appeal to data scientists, ML practitioners, and deep learning enthusiasts who are looking to perform accelerated, browser-based ML on the web using TensorFlow.js. Working knowledge of the JavaScript programming language is all you need to get started.

What this book covers

Chapter 1, Machine Learning for the Web, will show you the importance of ML on the web platform. Fundamentally, ML applications should provide some value to the users through a user-facing interface such as a web platform. In this chapter, we will leverage ML on the web platform to remove the fences between the user-facing environment and the environment where traditional server-side ML runs. You will learn how to install TensorFlow.js and set up the environment around it.

Chapter 2, Importing Pretrained Models into TensorFlow.js, explains how to import Keras pretrained models into TensorFlow.js. Since TensorFlow Core can train such a model efficiently, we can easily reuse the model in a client-side application.

Chapter 3, TensorFlow.js Ecosystem, shows you how to use some frameworks and libraries running with TensorFlow.js that are used to construct ML models, so that you can develop your own application more efficiently.

Chapter 4, Polynomial Regression, shows you how TensorFlow.js APIs are used with the simplest models. The application we look at predicts the y value of a sine curve with a given x value by using a polynomial regression model, implemented with a neural network.

Chapter 5, Classification with Logistic Regression, teaches you how to implement a classification model such as a logistic regression model. With the help of a practical example, we will teach you how to write a logistic regression application to classify flower types with the Iris dataset. 

Chapter 6, Unsupervised Learning, demonstrates the potential of TensorFlow as an ML framework by implementing a clustering algorithm such as k-means and demonstrating unsupervised learning. We will be implementing the k-means algorithm using the Iris dataset.

Chapter 7, Sequential Data Analysis, explains how the FFT algorithm is implemented in TensorFlow and how to use it in an ML application. You will also learn how complex numerical types are implemented in TensorFlow.js.

Chapter 8, Dimensionality Reduction, introduces t-SNE and how it can be implemented in TensorFlow.js.

Chapter 9, Solving Markov Decision Problems, introduces the implementation of the Bellman equation for solving MDP problems and explains how it is related to reinforcement learning.

Chapter 10, Deploying Machine Learning Applications, shows you the general ways to create a package from a TensorFlow.js application.

Chapter 11, Tuning Applications to Achieve High Performance, shows you how to make use of certain backend implementations to pursue high performance as well as giving you tips for tuning an application written in TensorFlow.js.

Chapter 12, Future Works around TensorFlow.js, covers more advanced features and optimizations implemented in TensorFlow.js so that you can learn about what is going on in TensorFlow.js projects.

To get the most out of this book

Working knowledge of JavaScript will be handy.

Download the example code files

You can download the example code files for this book from your account at www.packt.com. If you purchased this book elsewhere, you can visit www.packtpub.com/support and register to have the files emailed directly to you.

You can download the code files by following these steps:

Log in or register at www.packt.com.

Select the Support tab.

Click on Code Downloads.

Enter the name of the book in the Search box and follow the onscreen instructions.

Once the file is downloaded, please make sure that you unzip or extract the folder using the latest version of:

WinRAR/7-Zip for Windows

Zipeg/iZip/UnRarX for Mac

7-Zip/PeaZip for Linux

The code bundle for the book is also hosted on GitHub at https://github.com/PacktPublishing/Hands-On-Machine-Learning-with-TensorFlow.js. In case there's an update to the code, it will be updated on the existing GitHub repository.

We also have other code bundles from our rich catalog of books and videos available at https://github.com/PacktPublishing/. Check them out!

Download the color images

We also provide a PDF file that has color images of the screenshots/diagrams used in this book. You can download it here: http://www.packtpub.com/sites/default/files/downloads/9781838821739_ColorImages.pdf.

Conventions used

There are a number of text conventions used throughout this book.

CodeInText: Indicates code words in text, database table names, folder names, filenames, file extensions, pathnames, dummy URLs, user input, and Twitter handles. Here is an example: "GraphDef is a definition to describe the graph structure in the protocol buffer."

A block of code is set as follows:

message GraphDef {

  repeated NodeDef node = 1;

  VersionDef versions = 4;

  FunctionDefLibrary library = 2;

}

Any command-line input or output is written as follows:

$ ls my_tfjs_model

Bold: Indicates a new term, an important word, or words that you see on screen. For example, words in menus or dialog boxes appear in the text like this. Here is an example: "For example, ResNet-50, which is one of the major deep learning models for image classification, takes up 100 MB."

Warnings or important notes appear like this.

Tips and tricks appear like this.

Get in touch

Feedback from our readers is always welcome.

General feedback: If you have questions about any aspect of this book, mention the book title in the subject of your message and email us at [email protected].

Errata: Although we have taken every care to ensure the accuracy of our content, mistakes do happen. If you have found a mistake in this book, we would be grateful if you would report this to us. Please visit www.packtpub.com/support/errata, selecting your book, clicking on the Errata Submission Form link, and entering the details.

Piracy: If you come across any illegal copies of our works in any form on the internet, we would be grateful if you would provide us with the location address or website name. Please contact us at [email protected] with a link to the material.

If you are interested in becoming an author: If there is a topic that you have expertise in and you are interested in either writing or contributing to a book, please visit authors.packtpub.com.

Bewertungen

Please leave a review. Once you have read and used this book, why not leave a review on the site that you purchased it from? Potential readers can then see and use your unbiased opinion to make purchase decisions, we at Packt can understand what you think about our products, and our authors can see your feedback on their book. Thank you!

For more information about Packt, please visit packt.com.

Section 1: The Rationale of Machine Learning and the Usage of TensorFlow.js

In this section, readers will explore how machine learning applications work on the web platform. They will also learn how to set up an environment to run TensorFlow.js. Furthermore, readers will learn how to import pretrained models from Keras into TensorFlow.js. This section will also cover the ecosystem around TensorFlow.js.

This section contains the following chapters:

Chapter 1, Machine Learning for the Web

Chapter 2, Importing Pretrained Models into TensorFlow.js

Chapter 3, TensorFlow.js Ecosystem

Machine Learning for the Web

In this book, we will learn how to use TensorFlow.js to create machine learning applications. You'll need to be familiar with the following in order to get started:

Web-based programming languages, such as JavaScript and TypeScript

Web platform technology stacks (only a basic knowledge is required)

The fundamentals of machine learning algorithms

In this chapter, we are going to clarify why machine learning on the web is crucial in modern machine learning use cases and when to use web technology so that you can run your applications. You will also be introduced to the basic APIs of TensorFlow.js so that you can construct machine learning models. These topics act as the basis for the chapters that follow.

In this chapter, we will cover the following topics:

Why machine learning on the web?

Operation graphs

What is TensorFlow.js?

Installing TensorFlow.js

The low-level API

The Layers API

Technical requirements

In this chapter, as a prerequisite, you need to prepare the following libraries or frameworks in your