BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:From distributional ambiguity to gradient flows: Wasserstein\, Fis
 her-Rao\, and kernel approximation
DTSTART:20241128T140000
DTEND:20241128T160000
DTSTAMP:20260509T232027Z
UID:146d573a90849a229823ca09c2a2c67870ccf1d19eafba968945a212
CATEGORIES:Conferences - Seminars
DESCRIPTION:Dr. Jia-Jie Zhu\nWeierstrass Institute\, Berlin\nAbstract: Rec
 ent advances in distributionally robust optimization are distinct in that 
 their ambiguity sets are motivated by the theory of optimal transport and 
 information divergences. The theoretical foundation of those fields has re
 ceived a major push from the theory of PDE and gradient flows over the las
 t couple of decades. Motivated by several applications in inference and ge
 nerative models\, I will provide a few new results regarding the kernel ap
 proximation of Wasserstein and Fisher-Rao gradient flows\, such as a hidde
 n link between the flows of kernel maximum-mean discrepancy and relative e
 ntropies. These findings not only advance our theoretical understanding bu
 t also provide practical tools for enhancing machine learning algorithms. 
 Finally\, I will showcase inference and sampling algorithms using a new ke
 rnel approximation of the Wasserstein-Fisher-Rao (a.k.a. Hellinger-Kantoro
 vich) gradient flows\, which have better convergence characterization and 
 improved performance in computation.\n\nBio sketch: Jia-Jie Zhu is a machi
 ne learner\, applied mathematician\, and research group leader at the Weie
 rstrass Institute\, Berlin. Previously\, he worked as a postdoctoral resea
 rcher in machine learning at the Max-Planck-Institute for Intelligent Syst
 ems\, Tübingen\, and received his Ph.D. training in optimization\, at the
  University of Florida\, USA. He is interested in the intersection of mach
 ine learning\, analysis\, and optimization\, on topics such as gradient fl
 ows of probability measures\, optimal transport\, and robustness of learni
 ng and optimization algorithms.\n\n\n\n 
LOCATION:ODY 4 03 https://plan.epfl.ch/?room==ODY%204%2003
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR