Starts 26 Jun 2025 10:00
Ends 26 Jun 2025 11:00
Central European Time
QLS Seminar - An introduction to diffusion models in generative machine learning
ICTP
Common Area Old SISSA building Second floor
Via Beirut, 2
Abstract:
Diffusion models are a class of generative models in machine learning that iteratively transform data into noise through a forward diffusion process, typically converging toward a Gaussian distribution. A corresponding reverse process then reconstructs data-like samples by denoising these vectors step-by-step. This pair of processes resembles coarse-graining and fine-graining operations. In this talk, I will provide a brief introduction to the principles and mechanics of diffusion models, focusing on how they generate realistic samples from noise.
Diffusion models are a class of generative models in machine learning that iteratively transform data into noise through a forward diffusion process, typically converging toward a Gaussian distribution. A corresponding reverse process then reconstructs data-like samples by denoising these vectors step-by-step. This pair of processes resembles coarse-graining and fine-graining operations. In this talk, I will provide a brief introduction to the principles and mechanics of diffusion models, focusing on how they generate realistic samples from noise.