Naram Mhaisen and Georgios Iosifidis: Optimistic Learning at TU Delft, Foundational AI at Shelland Applications to Resource Management
03 September 2024 16:00 till 17:00 - Location: ECHO, building 29.02 215, Meeting Room A - By: ELLIS Delft | Add to my calendar
by Naram Mhaisen and Georgios Iosifidis | Delft University of Technology
Abstract
This talk addresses one of the most general frameworks for modeling sequential decision problems: online learning via (non-)convex optimization. We focus on a recent and practical twist on online learning, which is optimism. Optimistic learning revolves around the incorporation of untrusted machine learning advice into the learning process. The aim is simple yet ambitious: to boost learning performance when this advice proves accurate while ensuring a fallback to a desirable performance guarantee should they fail. The talk will present current results on this thread from our recent work, including the development of optimistic learning algorithms for continuous, discrete, and stateful environments. We will also present use cases where this framework finds practical applications, particularly in the fields of network optimization and control.
More information:
https://www.futurenetworkslab.net
https://github.com/Naram-m
This meeting is open for all interested researchers, and we particularly want to emphasize that we very much welcome all the PhD students and postdocs that are associated with the unit! If you have ideas about ways to make these talks more engaging for you, please let us know your suggestions.