O'Reilly logo
live online training icon Live Online training

Hands-on With Google Cloud AutoML

Build High-Quality Custom ML Models With Minimal Effort

Nitya Narasimhan

Machine learning and artificial intelligence are rapidly permeating all aspects of the technology ecosystem across enterprise and consumer domains. Thanks to the power of cloud computing, we now have machines that are trained to see (computer vision), talk (speech), understand (natural language processing) and even translate across different languages.

App developers can take advantage of pre-trained models for these features from cloud platform vendors like Google, integrating them using REST APIs exposed by those cloud ML services.But what if you wanted to have a custom model that was tailored more precisely to the needs and context for your domain -- but weren’t comfortable enough with the data sciences knowledge required to work with TensorFlow or other ML frameworks?

AutoML can help. In this training, we’ll learn what AutoML is, and how we can use Google’s Cloud AutoML products to train custom models for Natural Language (to classify documents), Translation (to interpret queries) and Vision (to label images) with minimal effort.

What you'll learn-and how you can apply it

  • Understand ideas behind AutoML like transfer learning and learning2learn
  • Learn about Google’s Cloud AutoML products (Vision, Natural Language, Translation)
  • Walk through the usage of each product with an example application and dataset
  • Do deep dives into key aspects like data preparation & model evaluation
  • Learn how AutoML differs from Cloud ML, TensorFlow & Firebase ML Kit options

This training course is for you because...

  • You are a technologist interested in understand cutting-edge technologies and best practices for machine learning and AI
  • You’ve heard of multiple Google ML technologies (Cloud ML, Auto ML, TensorFlow, Firebase ML Kit) and want to understand how and where they could be used
  • You’re a developer interested in integrating computer vision, natural language or translation capabilities into your application
  • You’re interested in exploring custom models because existing generic cloud ML solutions don’t offer sufficient accuracy or flexibility for your needs
  • You have access to large data sets (images, text) and want to see how AutoML can be used to create new knowledge or experiences for your product

Prerequisites

  • Basic familiarity with Machine Learning terminology
  • Some understanding of Google Cloud Platform (or similar) services
  • Development machine with browser (preferable Chrome) and CLI (terminal)

Training will involve using GCP console (for training) and CLI (for testing) custom models. We won’t write code but will work with example datasets provided by the instructor. Participants can bring their own datasets to explore independently but we will limit discussion to the Q&A segments in order to stay on time.

Course Set-up - Setup a Google Cloud Platform (GCP) account - Create a GCP project and enable billing on it - Training Website: https://automl18.bitnbot.com

Recommended Preparation

TensorFlow, Machine Learning and Learning to Learn (O’Reilly AI Conf)

Getting Started With Google Cloud Platform (Chad Smith)

Essential Machine Learning and AI (Noah Smith)

Recommended Follow-up

Deep Learning with TensorFlow (Jon Krohn)

Deep Learning for Natural Language Processing (Jon Krohn)

Pragmatic AI: An Introduction to Cloud Based Machine Learning (Noah Gift)

About your instructor

  • Nitya Narasimhan is a PhD with 20+ years of software development & research experience in distributed systems, mobile & web computing. She manages the Google Developer Group New York City (GDG NYC) chapter, organizes the DevFest NYC conference and speaks regularly on emerging technologies, cross-platform application development, machine learning and community. She is also a Google Developer Expert in Flutter and a technology educator and consultant based in New York.

Schedule

The timeframes are only estimates and may vary according to how the class is progressing

Segment 1 Introduction to AutoML (25 minutes)

  • What is the vision behind AutoML?
  • High level overview of Learning2Learn, Transfer Learning
  • Where does AutoML fit into Google’s Machine Learning ecosystem?
  • What is Cloud AutoML?

Q&A Break (5 mins)

Segment 2 AutoML Vision (45 minutes)

  • Challenge: Labeling Images accurately
  • Your Options: Vision API vs. AutoML Vision
  • AutoML Vision Training (walkthrough)
  • Recap: Data Preparation
  • Recap: Model Evaluation
  • Recap: Model Testing

Q&A Break (5 mins)

Segment 3 AutoML Natural Language (35 minutes)

  • Challenge: Detect Entities & Sentiment in Conversation
  • Your Options: NL API vs. AutoML Natural Language
  • AutoML Natural Language Training (walkthrough)
  • Recap: Data Preparation
  • Recap: Model Evaluation
  • Recap: Model Testing

Q&A Break (5 mins)

Segment 4 AutoML Translation (25 minutes)

  • Challenge: Translate query language
  • Pick your option: Translation API vs. AutoML Translate
  • AutoML Translate Training (walkthrough)
  • Recap: Data Preparation
  • Recap: Model Evaluation
  • Recap: Model Testing

Q&A Break (5 mins)

Segment 5 Putting It All Together (25 minutes)

  • The Google ML Ecosystem: Cloud ML, Auto ML, TensorFlow, Firebase ML Kit
  • What We Saw: Auto ML Vision, NL, Translation
  • Where is it useful: Application Examples
  • What are the challenges: Preparation & Evaluation
  • Where next: Resources & Recommendations