Generative Modeling by Estimating Gradients of the Data Distribution - Stefano Ermon
Vložit
- čas přidán 21. 07. 2024
- Seminar on Theoretical Machine Learning
Topic: Generative Modeling by Estimating Gradients of the Data Distribution
Speaker: Stefano Ermon
Affiliation: Stanford University
Date: May 12, 2020
For more video please visit video.ias.edu
How is the dlog_p(x) used in the cost function and reverse diffusion at the same time?
is the slide available?
instead of using Lagevin, is it possible to do HMC for score-based sampling? If so are the pitfalls avoided?
WOOOOOOOOOOOOOOOOOOOOOOOOOOOOO
Who disliked this?
Langevin dynamics? More like Langewin Dynamics