Machine Learning in Elixir

Learning to Learn with Nx and Axon

by Sean Moriarity

Stable Diffusion, ChatGPT, Whisper—these are just a few examples of incredible applications powered by developments in machine learning. Despite the ubiquity of machine learning applications running in production, there are only a few viable language choices for data science and machine learning tasks. Elixir’s Nx project seeks to change that. With Nx, you can leverage the power of machine learning in your applications, using the battle-tested Erlang VM in a pragmatic language like Elixir. In this book, you’ll learn how to leverage Elixir and the Nx ecosystem to solve real-world problems in computer vision, natural language processing, and more.

The Elixir Nx project aims to make machine learning possible without the need to leave Elixir for solutions in other languages. And even if concepts like linear models and logistic regression are new to you, you’ll be using them and much more to solve real-world problems in no time.

Start with the basics of the Nx programming paradigm—how it differs from the Elixir programming style you’re used to and how it enables you to write machine learning algorithms. Use your understanding of this paradigm to implement foundational machine learning algorithms from scratch. Go deeper and discover the power of deep learning with Axon. Unlock the power of Elixir and learn how to build and deploy machine learning models and pipelines anywhere. Learn how to analyze, visualize, and explain your data and models.

Discover how to use machine learning to solve diverse problems from image recognition to content recommendation—all in your favorite programming language.

• Errata, typos, suggestions

• Source Code (zip file)

Releases:

• Preface
• Foundations of Machine Learning
  • Make Machines That Learn
    • Classifying Flowers
    • Learning with Elixir
    • Wrapping Up
  • Get Comfortable with Nx
    • Thinking in Tensors excerpt
    • Using Nx Operations
    • Representing the World
    • Going from def to defn
    • Wrapping Up
  • Harness the Power of Math
    • Understanding Machine Learning Math
    • Speaking the Language of Data
    • Thinking Probabilistically
    • Tracking Change
    • Wrapping Up
  • Optimize Everything
    • Learning with Optimization
    • Regularizing to Generalize
    • Descending Gradients
    • Peering into the Black Box
    • Wrapping Up
  • Traditional Machine Learning
    • Learning Linearly
    • Learning from Your Surroundings
    • Using Clustering
    • Making Decisions
    • Wrapping Up
• Deep Learning
  • Go Deep with Axon
    • Understanding the Need for Deep Learning
    • Breaking Down a Neural Network
    • Creating Neural Networks with Axon
    • Wrapping Up
  • Learn to See
    • Identifying Cats and Dogs
    • Introducing Convolutional Neural Networks
    • Improving the Training Process
    • Going Beyond Image Classification
    • Wrapping Up
  • Stop Reinventing the Wheel
    • Identifying Cats and Dogs Again
    • Fine-Tuning Your Model
    • Understanding Transfer Learning
    • Taking Advantage of the Machine Learning Ecosystem
    • Wrapping Up
  • Understand Text
    • Classifying Movie Reviews
    • Introducing Recurrent Neural Networks
    • Understanding Recurrent Neural Networks
    • Wrapping Up
  • Forecast the Future
    • Predicting Stock Prices
    • Using CNNs for Single-Step Prediction
    • Using RNNs for Time-Series Prediction
    • Tempering Expectations
    • Wrapping Up
  • Model Everything with Transformers
    • Paying Attention
    • Going from RNNs to Transformers
    • Using Transformers with Bumblebee
    • Wrapping Up
  • Learn Without Supervision
    • Compressing Data with Autoencoders
    • Learning a Structured Latent
    • Generating with GANs
    • Learning Without Supervision in Practice
    • Wrapping Up
• Machine Learning in Practice
  • Put Machine Learning into Practice
    • Deciding to Use Machine Learning
    • Setting Up the Application
    • Integrating Nx with Phoenix
    • Seeding Your Databases
    • Building the Search LiveView
    • Wrapping Up
  • That’s a Wrap
    • Learning from Experience
    • Diffusing Innovation
    • Talking to Large Language Models
    • Compressing Knowledge
    • Moving Forward