BEGIN:VCALENDAR VERSION:2.0 PRODID:researchseminars.org CALSCALE:GREGORIAN X-WR-CALNAME:researchseminars.org BEGIN:VEVENT SUMMARY:Kelin Xia/夏克林 (新加坡南洋理工大学) DTSTART:20201228T070000Z DTEND:20201228T074500Z DTSTAMP:20260423T024749Z UID:iccm2020/55 DESCRIPTION:Title: Persistent representations based deep learning for drug design\nby K elin Xia/夏克林 (新加坡南洋理工大学) as part of ICCM 2020\n\n\ nAbstract\nEffective molecular representation is key to the success of mac hine learning models for molecular data analysis. In this talk\, we will d iscuss a series of persistent representations\, including persistent homol ogy\, persistent spectral models\, and persistent Ricci curvature and thei r combination with deep learning models. Unlike traditional graph and netw ork models\, these filtration-induced persistent models can characterize t he multiscale topological and geometric information\, thus significantly r educe molecular data complexity and dimensionality. Feature vectors are o btained from various persistent attributes derived from topological and ge ometric invariants\, such as homology\, cohomology\, eigenvalues\, and Ric ci curvature. They are inputted into deep learning models\, in particular\ , random forest\, gradient boosting tree and convolutional neural network (CNN). Our persistent representations based molecular fingerprints can sig nificantly boost the performance of learning models in drug design.\n LOCATION:https://researchseminars.org/talk/iccm2020/55/ END:VEVENT END:VCALENDAR