This repository contains a collection of educational resources that we have prepared for teaching the basics of machine learning to various audiences. The goal is to have simple and accessible resources that can enable anyone, including those with no machine learning background, to engage and learn from these tutorials.
Our aim is to contribute to democratisation of machine learning by providing accessible educational resources to inspire everyone.
These tutorials are presented as notebooks that can be launched via Google Colab.
Videogames and artificial intelligence have a long and happy history together. In this tutorial we will play a familiar (and hard!) game, learn what an agent is, and how we can make them learn to play the game with superhuman abilities. The agents we will build are evaluated on their performance on the game, and selected with variation in a virtual circle of life inspired by evolution.
This tutorial shows how we can build important artificial intelligence models for natural language processing called language models. Language models are surprisingly effective models that learn to predict the next letter (or word) given previous letters (or words) --- they essentially learn which letters (or words) go well together. In this tutorial we teach you how the computer represents and processes language, and show how we can use a big chunk of text to learn language models and apply them on two tasks: decoding secret messages and generating text.
Creativity is central to human intelligence. In this tutorial we see where Artificial Intelligence (AI) meets creativity. We will show you how an AI can produce realistic looking images of everyday objects as well as works-of-art and imagine "spider-dogs". We refer to an AI that is able to do any one of these things as Generative Models and towards the end of this tutorial you will build your own one.
This tutorial explores the use of machine learning for solving the protein structure prediction problem. Although the tutorial does not actually solve the problem itself, it provides students with very basic background in coding and biology to get started, and trains their intuition on machine learning methods, with the help of visualisation and a few examples of folding simple protein structures.
This tutorial introduces students to a simple reinforcement learning (RL) setup used in research. It involves running pre-existing code to set up an RL environment and visualise it. Students can then look at how a completely random (untrained) agent behaves in these environments. We also include simple code that implements a reinforcement learning method that can train the agent to solve these simple tasks. The behaviour of the trained agent can be then visualised together with plots of how the agent evolves through training.
This tutorial teaches you the basic ideas that underline scientific thinking. We cover developing and testing new knowledge with the scientific method through experimentation and validation, while showing common pitfalls in the process. Through a series of games you will play as an agent trying to understand the world, you will get insight into some of the core ideas behind scientific thinking.
If you have any feedback, or would like to get in touch with us,
please reach out by opening a new issue on the GitHub repo or emailing us at
educational@deepmind.com.
This is not an officially supported Google product.