Hands-On Artificial Intelligence on Google Cloud Platform: Build intelligent applications powered by TensorFlow, Cloud AutoML, BigQuery, and Dialogflow

Cover
Title Page
About Packt
Copyright and Credits
Contributors
Table of Contents
Preface
Section 1: Basics of Google Cloud Platform
Chapter 1: Overview of AI and GCP
	Understanding the Cloud First strategy for advanced data analytics
		Advantages of a Cloud First strategy
		Anti-patterns of the Cloud First strategy 
	Google data centers
	Overview of GCP
	AI building blocks
		Data
			Storage
			Processing
			Actions
		Natural language processing 
		Speech recognition
		Machine vision
		Information processing and reasoning
		Planning and exploring
		Handling and control
		Navigation and movement
		Speech generation
		Image generation
	AI tools available on GCP
		Sight
		Language
		Conversation
	Summary
Chapter 2: Computing and Processing Using GCP Components
	Understanding the compute options
		Compute Engine
			Compute Engine and AI applications
		App Engine
			App Engine and AI applications
		Cloud Functions
			Cloud Functions and AI applications
		Kubernetes Engine
			Kubernetes Engine and AI applications
	Diving into the storage options
		Cloud Storage
			Cloud Storage and AI applications
		Cloud Bigtable
			Cloud Bigtable and AI applications
		Cloud Datastore
			Cloud Datastore and AI applications
		Cloud Firestore
			Cloud Firestore and AI applications
		Cloud SQL
			Cloud SQL and AI applications
		Cloud Spanner
			Cloud Spanner and AI applications
		Cloud Memorystore
			Cloud Memorystore and AI applications
		Cloud Filestore
			Cloud Filestore and AI applications
	Understanding the processing options
		BigQuery
			BigQuery and AI applications
		Cloud Dataproc
			Cloud Dataproc and AI applications
		Cloud Dataflow
			Cloud Dataflow and AI applications
	Building an ML pipeline 
		Understanding the flow design
		Loading data into Cloud Storage
		Loading data to BigQuery
		Training the model
		Evaluating the model
		Testing the model
	Summary
Section 2: Artificial Intelligence with Google Cloud Platform
Chapter 3: Machine Learning Applications with XGBoost
	Overview of the XGBoost library
		Ensemble learning
			How does ensemble learning decide on the optimal predictive model?
				Reducible errors – bias
				Reducible errors – variance
				Irreducible errors
				Total error
			Gradient boosting
			eXtreme Gradient Boosting (XGBoost)
	Training and storing XGBoost machine learning models
	Using XGBoost trained models
	Building a recommendation system using the XGBoost library
		Creating and testing the XGBoost recommendation system model 
	Summary
Chapter 4: Using Cloud AutoML
	Overview of Cloud AutoML 
		The workings of AutoML
		AutoML API overview
			REST source – pointing to model locations
			REST source – for evaluating the model
			REST source – the operations API
	Document classification using AutoML Natural Language
		The traditional machine learning approach for document classification
		Document classification with AutoML
			Navigating to the AutoML Natural Language interface
			Creating the dataset
			Labeling the training data
			Training the model
			Evaluating the model
				The command line
				Python
				Java
				Node.js
			Using the model for predictions
				The web interface
				A REST API for model predictions
				Python code for model predictions
	Image classification using AutoML Vision APIs
		Image classification steps with AutoML Vision 
			Collecting training images
				Creating a dataset
			Labeling and uploading training images
			Training the model
			Evaluating the model
				The command-line interface
				Python code
			Testing the model
				Python code
	Performing speech-to-text conversion using the Speech-to-Text API
		Synchronous requests
		Asynchronous requests
		Streaming requests
	Sentiment analysis using AutoML Natural Language APIs
	Summary
Chapter 5: Building a Big Data Cloud Machine Learning Engine
	Understanding ML
	Understanding how to use Cloud Machine Learning Engine
		Google Cloud AI Platform Notebooks
			Google AI Platform deep learning images
			Creating Google Platform AI Notebooks
			Using Google Platform AI Notebooks
			Automating AI Notebooks execution
	Overview of the Keras framework 
	Training your model using the Keras framework
	Training your model using Google AI Platform
	Asynchronous batch prediction using Cloud Machine Learning Engine
	Real-time prediction using Cloud Machine Learning Engine
	Summary
Chapter 6: Smart Conversational Applications Using DialogFlow
	Introduction to DialogFlow
		Understanding the building blocks of DialogFlow
	Building a DialogFlow agent
		Use cases supported by DialogFlow
	Performing audio sentiment analysis using DialogFlow
	Summary
Section 3: TensorFlow on Google Cloud Platform
Chapter 7: Understanding Cloud TPUs
	Introducing Cloud TPUs and their organization
		Advantages of using TPUs
	Mapping of software and hardware architecture
		Available TPU versions
		Performance benefits of TPU v3 over TPU v2
		Available TPU configurations
		Software architecture
	Best practices of model development using TPUs
		Guiding principles for model development on a TPU
	Training your model using TPUEstimator
		Standard TensorFlow Estimator API
		TPUEstimator programming model
		TPUEstimator concepts
		Converting from TensorFlow Estimator to TPUEstimator
	Setting up TensorBoard for analyzing TPU performance
	Performance guide
		XLA compiler performance
		Consequences of tiling
		Fusion
	Understanding preemptible TPUs
		Steps for creating a preemptible TPU from the console
		Preemptible TPU pricing
		Preemptible TPU detection 
	Summary
Chapter 8: Implementing TensorFlow Models Using Cloud ML Engine
	Understanding the components of Cloud ML Engine
		Training service
			Using the built-in algorithms
			Using a custom training application
		Prediction service
		Notebooks
		Data Labeling Service
		Deep learning containers
	Steps involved in training and utilizing a TensorFlow model
		Prerequisites
		Creating a TensorFlow application and running it locally
			Project structure recommendation
			Training data
	Packaging and deploying your training application in Cloud ML Engine
	Choosing the right compute options for your training job
		Choosing the hyperparameters for the training job
	Monitoring your TensorFlow training model jobs
	Summary
Chapter 9: Building Prediction Applications
	Overview of machine-based intelligent predictions
		Understanding the prediction process
	Maintaining models and their versions
	Taking a deep dive into saved models
		SignatureDef in the TensorFlow SavedModel
		TensorFlow SavedModel APIs
	Deploying the models on GCP
		Uploading saved models to a Google Cloud Storage bucket
		Testing machine learning models
		Deploying models and their version
	Model training example
	Performing prediction with service endpoints
	Summary
Section 4: Building Applications and Upcoming Features
Chapter 10: Building an AI application
	A step-by-step approach to developing AI applications
		Problem classification 
			Classification
			Regression
			Clustering
			Optimization
			Anomaly detection
			Ranking
			Data preparation
		Data acquisition 
		Data processing 
		Problem modeling 
		Validation and execution
			Holdout
			Cross-validation
			Model evaluation parameters (metrics)
			Classification metrics
		Model deployment
	Overview of the use case – automated invoice processing (AIP)
	Designing AIP with AI platform tools on GCP
		Performing optical character recognition using the Vision API
		Storing the invoice with Cloud SQL
			Creating a Cloud SQL instance
			Setting up the database and tables
			Enabling the Cloud SQL API 
			Enabling the Cloud Functions API 
			Creating a Cloud Function 
			Providing the Cloud SQL Admin role
		Validating the invoice with Cloud Functions
		Scheduling the invoice for the payment queue (pub/sub)
		Notifying the vendor and AP team about the payment completion
		Creating conversational interface for AIP
	Upcoming features
	Summary
Other Books You May Enjoy
Index