Agenda

16 Apr 2024 12:15

A connection between Tempering and Entropic Mirror Descent

Modalita telematica

Speaker: Francesca Romana Crucinio, King's College London

Link Zoom: https://unive.zoom.us/j/85153268624?pwd=MzBhdlA2M1B2dThJQ2Y5T0EwUE5PZz09

Abstract:
This talk explores the connections between tempering (for Sequential Monte Carlo; SMC) and entropic mirror descent to sample from a target probability distribution whose unnormalized density is known. We establish that tempering SMC corresponds to entropic mirror descent applied to the reverse Kullback-Leibler (KL) divergence and obtain convergence rates for the tempering iterates. Our result motivates the tempering iterates from an optimization point of view, showing that tempering can be seen as a descent scheme of the KL divergence with respect to the Fisher-Rao geometry, in contrast to Langevin dynamics that perform descent of the KL with respect to the Wasserstein-2 geometry. We exploit the connection between tempering and mirror descent iterates to justify common practices in SMC and derive adaptive tempering rules that improve over other alternative benchmarks in the literature.

Bio Sketch:
Francesca is a Lecturer in Statistics in the Department of Mathematics, King’s College London. She obtained her PhD in 2021 from the University of Warwick, under the supervision of Adam M. Johansen and Arnaud Doucet. Before joining King’s College, she worked as a postdoctoral researcher at the University of Warwick within the CoSInES project led by Gareth O. Roberts and at CREST, ENSAE, Institut Polytechnique de Paris with Nicolas Chopin.

Language

The event will be held in English

Organized by

Dipartimento di Scienze Ambientali, Informatica e Statistica - Statistica (Prosdocimi) [DESC-DAIS]

Search in the agenda