mit-deep-learning
Tutorials, assignments, and competitions for MIT Deep Learning related courses.
Hands-on Jupyter notebook tutorials from MIT's deep learning course covering neural networks, image segmentation, and AI image generation, all runnable in the browser via Google Colab.
This repository is a collection of hands-on tutorials created to accompany MIT's deep learning courses, led by Lex Fridman and a small team at MIT. The tutorials are written as interactive notebooks, which means you can open them in your browser and run the code step by step without installing anything on your computer, thanks to Google Colab support.
The tutorials cover several topics in machine learning. One introduces the basic building blocks of deep learning, including how neural networks take in data and make predictions. Another shows how an AI model can look at a video of driving and label every pixel as road, car, pedestrian, or another category. A third explores a type of AI called generative adversarial networks, where two models are trained against each other to produce realistic images.
There is also a competition called DeepTraffic, where participants train a simulated car to drive as fast as possible through highway traffic. That component links out to its own repository and website.
All the tutorials are paired with recorded lecture videos, written blog posts, and links to the underlying papers where relevant. The README itself is fairly short and the repository functions mainly as a companion to the MIT course materials rather than a standalone textbook.
Where it fits
- Run interactive deep learning exercises in your browser without installing any software, using Google Colab.
- Learn how neural networks make predictions by working through the step-by-step notebook on deep learning basics.
- Train a model to label every pixel in a driving scene as road, car, or pedestrian using the image segmentation tutorial.
- Experiment with generative adversarial networks to produce realistic images using the GAN tutorial notebook.