Tutorial on Probabilistic U-Nets

Using a Minimal PyTorch Implementation of the Hierarchical Probabilistic U-Net.

Part 1: Source Reconstruction	Part 2: Rescue the Randomness

🔗 Download Slides

📽️ Access Recording of the Talk Presented at the KITP Program on Building a Physical Understanding of Galaxy Evolution with Data-driven Astronomy

Complementary Resources

1. 1806.05034 Original Probabilistic U-Net Paper (Kohl et al. 2018) + GitHub Repo
2. 1905.13077 Original Hierarchical Probabilistic U-Net Paper (Kohl et al. 2019) + GitHub Repo
3. 1903.10145 Paper on KL Vanishing and Cyclical Beta Schedule (Fu, Li, et al. 2019)
4. 1810.00597 Paper on GECO Loss (Rezende and Viola 2018)
5. 2302.03026 Novel Method for Coverage Probability Test (Lemos et al. 2023)
6. Link to Blog Post on Understanding Variational Autoencoders (VAEs) (Joseph Rocca)
7. Link to Blog Post on Conditional Variational Autoencoders (Isaac Dykeman)