Steering Diffusion Models for Generative AI
Diffusion models are advancing generative AI across vision, scientific, and emerging natural language domains. As we scale data and compute, foundation models learn rich priors over high-dimensional and multimodal data. This talk focuses on leveraging these priors for solving complex downstream tasks using test-time scaling—with guidance and reinforcement learning—covering practical methods, trade-offs, and examples.
Date and Time
Location
Hosts
Registration
-
Add Event to Calendar
- Instituto Superior Técnico - Campus da Alameda
- Lisboa, Lisboa
- Portugal
- Building: Informática II
- Room Number: Room Prof. José Tribolet
- Contact Event Hosts
- Co-sponsored by Instituto Superior Técnico - Univ. Lisboa; IT; INESC-ID
Speakers
Morteza Mardani of NVIDIA
Biography:
Morteza Mardani is a Principal Scientist at NVIDIA Research leading generative AI. He is also a visiting researcher in the Electrical Eng. dept. at Stanford University. He previously served as a postdoctoral researcher and a research associate at Stanford from 2015 to 2020 and was a visiting scholar at UC Berkeley's RISE Lab in 2015. He earned his Ph.D. in Electrical Eng. from the University of Minnesota. His contributions to generative modeling and statistical learning have been recognized with several awards, including the IEEE Signal Processing Society Young Author Best Paper Award in 2017. He is an IEEE Senior Member, an elected Distinguished Industry Speaker for the IEEE Signal Processing Society, and serves on its Computational Imaging Technical Committee.