Hands-On Artificial Intelligence on Amazon Web Services: Decrease the time to market for AI and ML applications with the power of AWS

Hands-On Artificial Intelligence on Amazon Web Services

Decrease the time to market for AI and ML applications with the power of AWS

Subhashini Tripuraneni

Charles Song

BIRMINGHAM - MUMBAI

Hands-On Artificial Intelligence on Amazon Web Services

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 authors, 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: Pravin Dhandre

Acquisition Editor: Aditi Gour

Content Development Editor: Nazia Shaikh

Senior Editor: Ayaan Hoda

Technical Editor: Dinesh Chaudhary

Copy Editor: Safis Editing

Project Coordinator: Kirti Pisat

Proofreader: Safis Editing

Indexer: Priyanka Dhadke

Production Designer: Aparna Bhagat

First published: October 2019

Production reference: 1041019

Published by Packt Publishing Ltd.

Livery Place

35 Livery Street

Birmingham

B3 2PB, UK.

ISBN 978-1-78953-414-6

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 authors

Subhashini Tripuraneni has several years of experience leading AI initiatives in financial services and convenience retail. She has automated multiple business processes and helped to create a proactive competitive advantage for businesses via AI. She is also a seasoned data scientist, with hands-on experience building machine learning and deep learning models in a public cloud. She holds an MBA from Wharton Business School, with a specialization in business analytics, marketing and operations, and entrepreneurial management. In her spare time, she enjoys going to theme parks and spending time with her children. She currently lives in Dallas, TX, with her husband and children.

Charles Song is a solutions architect with a background in applied software engineering research. He is skilled in software development, architecture design, and machine learning, with a proven ability to utilize emerging technologies to devise innovative solutions. He has applied machine learning to many research and industry projects, and published peer-reviewed papers on the subject. He holds a PhD in computer science from the University of Maryland. He has taught several software engineering courses at the University of Maryland for close to a decade. In his spare time, he likes to relax in front of his planted aquariums, but also enjoys martial arts, cycling, and snowboarding. He currently resides in Bethesda, MD, with his wife.

About the reviewer

Doug Ortiz is an experienced enterprise cloud, big data, data analytics, and solutions architect who has architected, designed, developed, re-engineered, and integrated enterprise solutions. His other areas of expertise include Amazon Web Services, Azure, Google Cloud, business intelligence, Hadoop, Spark, NoSQL databases, and SharePoint, to name but a few. He is also the founder of Illustris, LLC.

Huge thanks to my wonderful wife, Milla, Maria, Nikolay, and our children, for all their support.

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 Artificial Intelligence on Amazon Web Services

About Packt

Why subscribe?

Contributors

About the authors

About the reviewer

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

Reviews

Section 1: Introduction and Anatomy of a Modern AI Application

Introduction to Artificial Intelligence on Amazon Web Services

Technical requirements

What is AI?

Applications of AI

Autonomous vehicles

AI in medical care

Personalized predictive keyboards

Why use Amazon Web Services for AI?

Overview of AWS AI offerings

Hands-on with AWS services

Creating your AWS account

Navigating through the AWS Management Console

Finding AWS services

Choosing the AWS region

Test driving the Amazon Rekognition service

Working with S3

Identity and Access Management

Getting familiar with the AWS CLI

Installing Python

Installing Python on macOS

Installing Python on Linux

Installing Python on Microsoft Windows

Windows 10

Earlier Windows versions

Installing the AWS CLI

Configuring the AWS CLI

Invoking the Rekognition service using the AWS CLI

Using Python for AI applications

Setting up a Python development environment

Setting up a Python virtual environment with Pipenv

Creating your first Python virtual environment

First project with the AWS SDK

Summary

References

Anatomy of a Modern AI Application

Technical requirements

Understanding the success factors of artificial intelligence applications

Understanding the architecture design principles for AI applications

Understanding the architecture of modern AI applications

Creation of custom AI capabilities

Working with a hands-on AI application architecture

Object detector architecture

Component interactions of the Object Detector

Creating the base project structure

Developing an AI application locally using AWS Chalice

Developing a demo application web user interface

Deploying AI application backends to AWS via Chalice

Deploying a static website via AWS S3

Summary

Further reading

Section 2: Building Applications with AWS AI Services

Detecting and Translating Text with Amazon Rekognition and Translate

Making the world smaller

Understanding the architecture of Pictorial Translator 

Component interactions of Pictorial Translator 

Setting up the project structure

Implementing services

Recognition service – text detection

Translation service – translating text

Storage service – uploading files

A recommendation on unit testing

Implementing RESTful endpoints

Translate the image text endpoint

Upload the image endpoint

Implementing the web user interface

index.html

scripts.js

Deploying Pictorial Translator to AWS

Discussing project enhancement ideas

Summary

Further reading

Performing Speech-to-Text and Vice Versa with Amazon Transcribe and Polly

Technical requirements

Technologies from science fiction

Understanding the architecture of Universal Translator

Component interactions of Universal Translator

Setting up the project structure

Implementing services

Transcription service – speech-to-text

Translation Service – translating text

Speech Service – text-to-speech

Storage Service – uploading and retrieving a file

Implementing RESTful endpoints

Translate recording endpoint

Synthesize speech endpoint

Upload recording Endpoint

Implementing the Web User Interface

index.html

scripts.js

Deploying the Universal Translator to AWS

Discussing the project enhancement ideas

Summary

References

Extracting Information from Text with Amazon Comprehend

Technical requirements

Working with your Artificial Intelligence coworker

Understanding the Contact Organizer architecture

Component interactions in Contact Organizer

Setting up the project structure

Implementing services

Recognition Service – text detection

Extraction Service – contact information extraction

Contact Store – save and retrieve contacts

Storage Service – uploading and retrieving a file

Implementing RESTful endpoints

Extract Image Information endpoint

Save contact and get all contacts endpoints

Upload image endpoint

Implementing the web user interface

Index.html

scripts.js

Deploying the Contact Organizer to AWS

Discussing the project enhancement ideas

Summary

Further reading

Building a Voice Chatbot with Amazon Lex

Understanding the friendly human-computer interface

Contact assistant architecture

Understanding the Amazon Lex development paradigm

Setting up the contact assistant bot

The LookupPhoneNumberByName intent

Sample utterances and slots for LookupPhoneNumberByName

Confirmation prompt and response for LookupPhoneNumberByName

Fulfillment for LookupPhoneNumberByName using AWS Lambda

DynamoDB IAM role for LookupPhoneNumberByName

Fulfillment lambda function for LookupPhoneNumberByName

Amazon Lex helper functions

The intent fulfillment for LookupPhoneNumberByName

Test conversations for LookupPhoneNumberByName

The MakePhoneCallByName intent

Sample utterances and lambda initialization/validation for MakePhoneCallByName

Slots and confirmation prompt for MakePhoneCallByName

Fulfillment and response for MakePhoneCallByName

Test conversations for MakePhoneCallByName

Deploying the contact assistant bot

Integrating the contact assistant into applications

Intelligent assistant service implementation

Contact assistant RESTful endpoint

Summary

Further reading

Section 3: Training Machine Learning Models with Amazon SageMaker

Working with Amazon SageMaker

Technical requirements

Preprocessing big data through Spark EMR

Conducting training in Amazon SageMaker

Learning how Object2Vec Works

Training the Object2Vec algorithm

Deploying the trained Object2Vec and running inference

Running hyperparameter optimization (HPO)

Understanding the SageMaker experimentation service

Bring your own model – SageMaker, MXNet, and Gluon

Bring your own container – R model

Summary

Further reading

Creating Machine Learning Inference Pipelines

Technical requirements

Understanding the architecture of the inference pipeline in SageMaker

Creating features using Amazon Glue and SparkML

Walking through the prerequisites

Preprocessing data using PySpark

Creating an AWS Glue job

Identifying topics by training NTM in SageMaker

Running online versus batch inferences in SageMaker

Creating real-time predictions through an inference pipeline

Creating batch predictions through an inference pipeline

Summary

Further reading

Discovering Topics in Text Collection

Technical requirements

Reviewing topic modeling techniques

Understanding how the Neural Topic Model works

Training NTM in SageMaker

Deploying the trained NTM model and running the inference

Summary

Further reading

Classifying Images Using Amazon SageMaker

Technical requirements

Walking through convolutional neural and residual networks

Classifying images through transfer learning in Amazon SageMaker

Creating input for image classification

Defining hyperparameters for image classification

Performing inference through Batch Transform

Summary

Further reading

Sales Forecasting with Deep Learning and Auto Regression

Technical requirements

Understanding traditional time series forecasting

Auto-Regressive Integrated Moving Average (ARIMA )

Exponential smoothing

How the DeepAR model works

Model architecture

Arriving at optimal network weights 

Understanding model sales through DeepAR

Brief description of the dataset

Exploratory data analysis

Data pre-processing

Training DeepAR

Predicting and evaluating sales

Summary

Further reading

Section 4: Machine Learning Model Monitoring and Governance

Model Accuracy Degradation and Feedback Loops

Technical requirements

Monitoring models for degraded performance

Developing a use case for evolving training data – ad-click conversion

Creating a machine learning feedback loop

Exploring data

Creating features

Using Amazon's SageMaker XGBoost algorithm to classify ad-click data

Evaluating model performance

Summary

Further reading

What Is Next?

Summarizing the concepts we learned in Part I

Summarizing the concepts we learned in Part II

Summarizing the concepts we learned in Part III

Summarizing the concepts we learned in Part IV

What's next?

Artificial intelligence in the physical world

AWS DeepLens

AWS DeepRacer

Internet of Things and AWS IoT Greengrass

Artificial intelligence in your own field

Summary

Other Books You May Enjoy

Leave a review - let other readers know what you think

Preface

Hands-On Artificial Intelligence on Amazon Web Services teaches you about the various artificial intelligence and machine learning services available on AWS. Through practical hands-on exercises, you'll learn how to use these services to generate impressive results. You will be able to design, develop, monitor, and maintain machine learning and deep learning models on AWS effectively.

The book starts with an introduction to AI and its applications in different industries along with an overview of AWS on AI/machine learning services and platforms. It will teach you all about detecting and translating text with Amazon Rekognition and Amazon Translate. You will learn how to perform speech-to-text with the help of Amazon Transcribe and Amazon Polly.

It covers the use of Amazon Comprehend for extracting information from text and Amazon Lex for building voice chatbots. You will gain an understanding of the key capabilities of Amazon SageMaker – wrangling big data, discovering topics in text collections, and classifying images. Lastly, the book explores sales forecasting with deep learning and auto regression and model accuracy degradation.

By the end of this book, you will have all the knowledge required to work with and implement AI in AWS through immersive hands-on exercises covering all the aspects of the model life cycle.

Who this book is for

This book is ideal for data scientists, machine learning developers, deep learning researchers, and AI enthusiasts who want to use the power of AWS services to implement powerful AI solutions. A basic understanding of machine learning concepts is expected.

What this book covers

Chapter 1, Introduction to Artificial Intelligence on Amazon Web Services, introduces the umbrella term AI, which includes machine learning and deep learning. We will cover some of the hottest topics in AI, including image recognition, natural language processing, and speech recognition. We will provide a high-level overview of AWS' AI and machine learning services and platforms. AWS offers both managed services for ready-to-use AI/machine learning capabilities and managed infrastructures to train your own custom machine learning models. We will provide guidance on when to leverage managed services and when to train custom machine learning models. You will learn how to install and configure your development environment. We will guide you through the process of setting up for Python, the AWS SDK, and web development tools you will need for the hands-on projects in later chapters. We will also help you to verify your environment setup with working code that interacts with AWS platforms programmatically.

Chapter 2, Anatomy of a Modern AI Application, dives into the architecture and components of a modern AI application. We start to introduce patterns and concepts of a well-architected application; these will help you to design production-grade intelligent solutions. Not only will these concepts help you to rapidly experiment with and prototype solutions, but they will also help you to develop solutions that are flexible, extensible, and maintainable throughout the application life cycle. You will build the skeleton of a target architecture, which you will fill out in later chapters.

Chapter 3, Detecting and Translating Text with Amazon Rekognition and Translate, demonstrates how to build your first AI application that can translate foreign texts appearing in pictures into their native language. You will get hands-on experience with Amazon Rekognition and Amazon Translate. You will first build a reusable framework with AI and machine learning capabilities from AWS, and then build the application on top of that framework. We will demonstrate how separation between the capabilities and application logic can lead to flexibility and reusability; a concept that will become increasingly clear as the hands-on projects continue in later chapters.

Chapter 4, Performing Speech-to-Text and Vice Versa with Amazon Transcribe and Polly, shows you how to build an application that can translate voice conversations to and from different languages. You will get hands-on experience with Amazon Transcribe and Amazon Polly. Not only will you continue to build out the reusable framework of AI capabilities, but you will also reuse the translation capability of the framework that you built in the previous chapter. This will reinforce the concepts and benefits of well-architected production-ready AI solutions to increase experimentation and speed to market.

Chapter 5, Extracting Information from Text with Amazon Comprehend, demonstrates how you can build an application that can extract and organize information from photos of business cards. You will gain hands-on experience of Amazon Comprehend and reuse the text detection capability from previous chapters. In addition to this, we will introduce the concept of human-in-the-loop. You will build a human-in-the-loop graphical user interface that can allow users to verify and even correct information extracted by Amazon Comprehend.

Chapter 6, Building a Voice Chatbot with AWS Lex, enables you to continue the hands-on project by building a voice chatbot to look up business card contact information that was extracted and stored in the previous project. You will get hands-on experience of building a chatbot with Amazon Lex and integrating the chatbot interface into the application as a digital assistant.

Chapter 7, Working with Amazon SageMaker, explores the key capabilities of Amazon SageMaker – from wrangling big data to training and deploying a built-in model (Object2Vec), to identifying the best performing model to bring your own model and container to the SageMaker ecosystem. We illustrate each of the components through the book ratings dataset. First, we predict the rating of a book for a given user; that is, a book that the user has never rated. Second, we automate hyperparameter optimization with SageMaker's HPO capability, while also discovering the best performing model and its corresponding train and test sets through SageMaker search. Third, we illustrate how seamless it is to bring your model and container to SageMaker, avoiding the effort required to rebuild the same model in SageMaker. By the end of this chapter, you will know how to leverage all the key features of Amazon SageMaker.

Chapter 8, Creating Machine Learning Inference Pipelines, walks you through how SageMaker and other AWS services can be employed to create machine learning pipelines that can process big data, train algorithms, deploy trained models, and run inferences—all while using the same logic for data processing during model training and inference.

Chapter 9, Discovering Topics in Text Collection, introduces a new topic. In all the preceding NLP chapters, you learned how to use several NLP services offered by Amazon. In order to have fine-grained control over the model training and deployment, and to build models for scale, we'll use algorithms in Amazon SageMaker.

Chapter 10, Classifying Images Using Amazon SageMaker, follows on from what you have learned about Amazon Rekognition. Here, you'll learn how to classify your own images beyond the predetermined images classified by the Rekognition API. In particular, we will focus on labeling our own image dataset and using SageMaker's image classification algorithm to detect custom images. We will learn how to conduct transfer learning from ResNet50, a pretrained deep residual learning model trained on ImageNet (an image database organized by nouns and supported by Stanford University and Princeton University).

Chapter 11, Sales Forecasting with Deep Learning and Autoregression, explains how deep learning and autoregression (DeepAR) can be used for sales forecasting. In particular, a thorough understanding will be gained of Long Short Term Memory (LSTM), a form of Recurrent Neural Networks (RNNs). RNNs are networks with loops, allowing information to persist, connecting previous information to the present task. Autoregression uses observations from previous time steps as input to the regression equation to predict the value at the next time step. By the end of this chapter, you will have built a robust sales forecasting model with Amazon SageMaker.

Chapter 12, Model Accuracy Degradation and Feedback Loops, explains why models degrade in production. To illustrate this, we discuss how to predict ad-click conversion for mobile apps. As new data becomes available, it is important to retrain models to achieve optimal production performance. 

Chapter 13, What Is Next?, summarizes the concepts that we have learned so far. Additionally, we will briefly discuss the AI frameworks and infrastructures that AWS offers.

To get the most out of this book

A basic understanding of machine learning and AWS concepts is expected.

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-Artificial-Intelligence-on-Amazon-Web-Services. 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: https://static.packt-cdn.com/downloads/9781789534146_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: Let's create a directory for the project and name it ObjectDetectionDemo.

A block of code is set as follows:

{

  Image: {

    Bytes: ...

    }

}

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

$ brew install python3

$ brew install pip3

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: "The capability is built using deep learning techniques such as automatic speech recognition (ASR) and natural language understanding (NLU) in order to convert speech into text and to recognize intents within text."

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.

Reviews

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: Introduction and Anatomy of a Modern AI Application

This section aims to introduce artificial intelligence (AI) and provide an overview of AI capabilities offered by Amazon Web Services (AWS). It will provide a step-by-step setup for AI development on AWS, including the AWS Software Development Kit (SDK) and Python development toolset. Additionally, it will give an introduction to the components and architecture of a modern AI application.

This section comprises the following chapters:

Chapter 1, Introduction to Artificial Intelligence on Amazon Web Services

Chapter 2, Anatomy of a Modern AI Application

Introduction to Artificial Intelligence on Amazon Web Services

In this chapter, we will start with a high-level overview of artificial intelligence (AI), including its history and the broad set of methods that it uses. Then, we will take a look at a few applications of AI that have the potential to profoundly change our world. With growing interests in AI, many companies, including Amazon, are offering a plethora of tools and services to help developers create intelligent-enabled applications. We will provide a high-level overview of AI offerings from Amazon Web Services, and we will also provide our guidance on how to best leverage them. Being a hands-on book, we will quickly dive into intelligent-enabled application development with Amazon Web Services.

We will cover the following topics:

Overview of AI and its applications.

Understanding the different types of Amazon Web Services offerings for AI.

How to set up an Amazon Web Servicesaccount and the environment for intelligent-enabled application development.

Get hands-on experience with Amazon Rekognition and other supporting services.

Develop our first intelligent-enabled application.

Technical requirements

This book's GitHub repository, which contains the source code for this chapter, can be found at https://github.com/PacktPublishing/Hands-On-Artificial-Intelligence-on-Amazon-Web-Services.

What is AI?

AI is an umbrella term that describes a branch of computer science that aims to create intelligent agents. The field of AI is highly technical and specialized; there is a broad set of theories, methods, and technologies in AI that allow computers to see (computer vision), to hear (speech recognition), to understand (natural language processing), to speak (text-to-speech), and to think (knowledge reasoning and planning).

It may seem that AI is a buzzword of our current times, but it has existed since the 1950s, when early work on artificial neural networks that mimic led the human brain stirred up excitement for thinking machines. With all the fanfare it receives in the media today, it is hard to believe that this field had to endure two AI winters, where interest in AI research and development dwindled. Today, AI has become popular again, thanks to the increased volume of data, cheaper storage, advancements in algorithms, and an increase in computing power.

One of the most important subfields of AI is machine learning (ML). ML is such a prominent part of AI that these two terms are often used interchangeably today. ML is the most promising set of techniques to achieve AI. These techniques gave us a new way to program computers through self-learning algorithms that can derive knowledge from data. We can train ML models that can look for patterns and draw conclusions like humans would. With these self-learning algorithms, the data itself has become the most valuable asset. Data has become the competitive advantage in industries; it is the new intellectual property. Between similar ML techniques (even inferior ML techniques), the best data will win.

What's old is