BEGIN:VCALENDAR VERSION:2.0 PRODID:researchseminars.org CALSCALE:GREGORIAN X-WR-CALNAME:researchseminars.org BEGIN:VEVENT SUMMARY:Justin M. Curry (University at Albany) DTSTART:20251110T130000Z DTEND:20251110T140000Z DTSTAMP:20260422T140215Z UID:BilTop/125 DESCRIPTION:Title: Stratification Theory for Reinforcement Learning\nby Justin M. Curry (University at Albany) as part of Bilkent Topology Seminar\n\nLecture held in SA 141.\n\nAbstract\nIn this talk I will use the framework of poset-st ratified spaces to study games\, where reward can be both discrete and con tinuous. Following work by Yuliy Baryshnikov\, I will show how certain vid eo games naturally give rise to stratified spaces. Surprisingly\, when mod ern neural nets are trained to play these same video games\, a similar str atification structure can be observed in their latent representations. Our methods follow recent work by Michael Robinson and others on using Volume Growth Laws to detect non-manifold structure in the token space for LLMs. We expand and strengthen Robinson’s analysis by considering non-textual data and prove a realization result for volume growth in a stratified spa ce. This is joint work with Brennan Lagasse\, Ngoc B. Lam\, Gregory Cox\, David Rosenbluth\, and Alberto Speranzon.\n LOCATION:https://researchseminars.org/talk/BilTop/125/ END:VEVENT END:VCALENDAR