JAX seminar and models

The notebooks from the Jax seminar can be divided into two groups: those that illustrate the basics of the JAX library and those that show the results of the models studied during the seminar. The model notebooks are listed below. Click on the button to access the notebooks and lecture notes.

1. Convolutional Classifier

This is a simple 13 layer convolutional neural network that classifies the MNIST dataset: mnist.ipynb

2. Variational Autoencoder

This notebook demonstrates the use of a generative model - variational autoencoder - to generate new images of handwritten digits: vae_mnist.ipynb

3. Denoising Diffusion Model

This notebook demonstrates the use of a different type of generative model - denoising diffusion model - to generate new images: diffusion.ipynb

4. Reinforcement Learning

Proximal Policy Optimization (PPO) is a policy gradient method that is used to train agents. The notebook trains an agent in the classical cart-pole environment: ppo.ipynb

5. Chaos

This notebook demonstrates the use of JAX to solve the Lorenz system of differential equations. The chaotic behaviour of the solutions is faithfully captured: lorenz.ipynb