This repository contains the labs for 6.S184/6.S975: Generative AI with Stochastic Differential Equations, as taught at MIT over IAP 2025. For questions and concerns, please email us at ezraerives@gmail.com and phold@mit.edu. Enjoy :)
- 1/22/25: Lab 1: Fix several typos in lab one.
- 1/27/25: Labs 1 + 2: Add clarification regarding truncated PDFs when submitting w/ Colab + Chrome.
- 1/27/25: Lab 2: Clarify numerical issues when sampling conditional SDE for large noise levels. Clarify differences between score parameterizations.
- 3/9/25: Lab 1: resolve timestep bug in Langevin dynamics (thanks Ádám!)
- 3/12/25: Lab 3: Fix guidance embedding dimension bug (thanks Roger!)
- 3/13/25: Lab 3 solutions: Fix bug when sampling conditioning variable (thanks Zewen!)
- 1/11/26: Lab 2: Fix "doubly stochastic" diffusion coefficient implementation in
ConditionalVectorFieldSDEandLangevinFlowSDE(thanks Mariia!)
We would like to thank the following individuals for their invaluable feedback and suggestions:
- Cameron Diao
- Tally Portnoi
- Andi Qu
- Ádám Burián
- Roger Trullo
- Zewen Yang
- Mariia Drozdova
