BEGIN:VCALENDAR VERSION:2.0 PRODID:researchseminars.org CALSCALE:GREGORIAN X-WR-CALNAME:researchseminars.org BEGIN:VEVENT SUMMARY:Carlos Floyd (University of Chicago) DTSTART:20250313T163000Z DTEND:20250313T170000Z DTSTAMP:20260421T125207Z UID:MoRN/119 DESCRIPTION:Title: L imits on the computational expressivity of non-equilibrium biophysical pro cesses\nby Carlos Floyd (University of Chicago) as part of Seminar on the Mathematics of Reaction Networks\n\n\nAbstract\nMany biological decisi on-making processes can be viewed as performing a classification task over a set of inputs\, using various chemical and physical processes as "biolo gical hardware." In this context\, it is important to understand the inher ent limitations on the computational expressivity of classification functi ons instantiated in biophysical media. Here\, we model biochemical network s as Markov jump processes and train them to perform classification tasks\ , allowing us to investigate their computational expressivity. We reveal s everal unanticipated limitations on the input-output functions of these sy stems\, which we further show can be lifted using biochemical mechanisms l ike promiscuous binding. We analyze the flexibility and sharpness of decis ion boundaries as well as the classification capacity of these networks. A dditionally\, we identify distinctive signatures of networks trained for c lassification\, including the emergence of correlated subsets of spanning trees and a creased "energy landscape" with multiple basins. Our findings have implications for understanding and designing physical computing syste ms in both biological and synthetic chemical settings.\n\nThis is joint wo rk with Suri Vaikuntanathan\, Arvind Murugan\, and Aaron Dinner (https://a rxiv.org/abs/2409.05827).\n LOCATION:https://researchseminars.org/talk/MoRN/119/ END:VEVENT END:VCALENDAR