BEGIN:VCALENDAR VERSION:2.0 PRODID:researchseminars.org CALSCALE:GREGORIAN X-WR-CALNAME:researchseminars.org BEGIN:VEVENT SUMMARY:Yulia R. Gel (UT Dallas - USA) DTSTART:20230818T160000Z DTEND:20230818T170000Z DTSTAMP:20260423T022926Z UID:GEOTOP-A/48 DESCRIPTION:Title: Coupling Time-Aware Multipersistence Knowledge Representation with Graph Convolutional Networks for Time Series Forecasting\nby Yulia R. Gel ( UT Dallas - USA) as part of GEOTOP-A seminar\n\n\nAbstract\nGraph Neural N etworks (GNNs) are proven to be a powerful machinery for learning complex dependencies in multivariate spatio-temporal processes. However\, most exi sting GNNs have inherently static architectures\, and as a result\, do not explicitly account for time dependencies of the encoded knowledge and are limited in their ability to simultaneously infer latent time-conditioned relations among entities. We postulate that such hidden time-conditioned p roperties may be captured by the tools of multipersistence\, i.e.\, an eme rging machinery in topological data analysis which allows us to quantify d ynamics of the data shape along multiple geometric dimensions. We propose to summarize inherent time-conditioned topological properties of the data as time-aware multipersistence Euler-Poincaré surface and prove its stabi lity. We then construct a supragraph convolution module which simultaneous ly accounts for the extracted intra- and inter-dependencies in the data. W e illustrate the utility of the proposed approach in application to foreca sting highway traffic flow\, blockchain Ethereum token prices\, and COVID- 19 hospitalizations.\n LOCATION:https://researchseminars.org/talk/GEOTOP-A/48/ END:VEVENT END:VCALENDAR