BEGIN:VCALENDAR
VERSION:2.0
PRODID:OpenCms ustutt.16.0.0.32
BEGIN:VTIMEZONE
TZID:Europe/Berlin
X-LIC-LOCATION:Europe/Berlin
BEGIN:DAYLIGHT
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
DTSTART:19700329T020000
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=3
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
DTSTART:19701025T030000
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=10
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20221202T164621
UID:780a908e-7258-11ed-8a72-000e0c3db68b
SUMMARY:Vortrag von Dr. Michael Mühlebach
DESCRIPTION:Dr. Michael Mühlebach\nMax Planck Institute for Intelligent Systems\nTübingen, Germany\n \nTuesday 2022-12-13 4 p.m.\nIST Seminar Room 2.255 - Pfaffenwaldring 9 - Campus Stuttgart-Vaihingen \nAbstract\nMy talk will highlight connections between dynamical systems and optimization and will be\ndivided into two parts: The first part presents an analysis of accelerated first-order optimization\nalgorithms, where the continuous dependence of iterates with respect to their initial conditions\nwill be exploited for characterizing the convergence rate. The result establishes criteria for\naccelerated convergence of a large class of momentum-based optimization algorithms. The criteria,\nwhich are easily verifiable, are necessary and sufficient and therefore precisely characterize\noptimization algorithms that are accelerated. The analysis applies to non-convex functions, unifies\ndiscrete-time and continuous-time models, and rigorously explains why structure-preserving\n(symplectic) discretization schemes are important in optimization. The second part of the talk\nintroduces a class of first-order methods for constrained optimization that are based on an analogy\nto non-smooth dynamical systems. The key underlying idea is to express constraints in terms of\nvelocities instead of positions, which has the algorithmic consequence that optimizations over\nfeasible sets at each iteration are replaced with optimizations over local, sparse convex\napproximations. The result is a simplified suite of algorithms and an expanded range of possible\napplications in machine learning. \nBiographical Information\nMichael Muehlebach studied mechanical engineering at ETH Zurich. He received his Ph.D. under the\nsupervision of Prof. R. D'Andrea in 2018 and joined the group of Prof. Michael I. Jordan at the\nUniversity of California, Berkeley as a postdoctoral researcher. In 2021 he started as an\nindependent group leader at the Max Planck Institute for Intelligent Systems in Tuebingen, where he\nleads the group "learning and dynamical systems".\nHe is interested in a variety of subjects, including machine learning, dynamical systems, and\noptimization. \nHe received the Outstanding D-MAVT Bachelor Award for his Bachelor's degree and the Willi-Studer\nprize for the best Master's degree. His Ph.D. thesis was awarded with the ETH Medal and the HILTI\nprize for innovative research. He was also awarded a Branco Weiss Fellowship and an Emmy Noether\nFellowship, which fund his research group.\n
DTSTART;VALUE=DATE:20221213
URL;VALUE=URI:https://www.ist.uni-stuttgart.de/de/veranstaltungen/Vortrag-von-Dr.-Michael-Muehlebach/
END:VEVENT
END:VCALENDAR