Generative Modeling

Outline

Shared with: Bruno Galerne
Timetable: Monday 9:30 to 12:30
All the information about the course can be found here. The image in the description of the course is extracted from the paper Score-Based Generative Modeling with Critically-Damped Langevin Diffusion .

Slides

  • First part of the course handled by Bruno Galerne, see here
  • Introduction to Generative Modeling
  • Introduction to Score-Based Generative Modeling
  • Theory of Score-Based Generative Modeling
  • Introduction to Schr\"odinger Bridge

    • Lab session

  • MNIST tutorial by Yang Song

    • Final exam

    The goal is to understand, reimplement (not necessarily in high-dimensional settings), discuss and possibly extend one of the following paper:
  • Tackling the Generative Learning Trilemma with Denoising Diffusion GANs
  • Score-Based Generative Modeling with Critically-Damped Langevin Diffusion
  • Autoregressive Diffusion Models
    • The students will team up (two students per team) and present their findings during an oral presentation (30 minutes). They will also write a report on the project.
    generative modeling M2 2021-2022