Plans for the week of April 21-25
Dear all, we hope you've had a great Easter break and were able to recharge properly.
And thx so much for all the nice reports (and progress reports as well) for project 1. It has been a pleasure to read them and grade those which were finalized as project 1.
Before the break we started discussing generative models and went from energy based models (with an emphasis on Boltzmann machines) to a presentation of the math of variational autoencoders (VAEs). This week will start with a reminder of the basic math of VAEs, and discuss how to implement VAEs using either tensorflow/keras and Pytorch. If we get time we will start discussing Diffusion models (which in a simplistic way can be viewed as stacked VAEs).
The plans for the rest of the semester are (tentative)
April 24: Math and implementation of VAEs, begin (if we get time) with diffusion models, project work
May 1: Public holiday, no lecture
May 8: Diffusion models, math and implementations, project work
May 15: Generalized adversarial networks (GANs) , summary of course and project work
May 22: if you are not fed up, we may either focus on project work and/or have a short intro to reinforcement learning
The plans for this week are thus:
Plans for the week April 21-25, 2025
Deep generative models.
-
Discussion of project 2
-
Variational Autoencoders (VAE), Mathematics and codes, continuation from last week
-
Reading recommendation:
a. Goodfellow et al sections 20.10-20-14 on VAEs
b. Calvin Luo https://calvinyluo.com/2022/08/26/diffusion-tutorial.html
c. An Introduction to Variational Autoencoders, by Kingma and Welling, see https://arxiv.org/abs/1906.02691
The jupyter-notebook is at https://github.com/CompPhysics/AdvancedMachineLearning/blob/main/doc/pub/week13/ipynb/week13.ipynb
Best wishes to you all,
Edvin and Morten