Schedule

Subscribe to our calendar for the most recent schedule.

May 21st 3:00pm CET - Taeyoung Yun

Title: Posterior Inference in Generative Models for High-dimensional Black-box Optimization

Abstract: The optimization of high-dimensional black-box functions is a ubiquitous challenge across many scientific and engineering disciplines. Traditional high-dimensional Bayesian optimization (BO) primarily relies on trust-region-based optimization of acquisition functions to select candidate points. In this seminar, I will introduce a novel alternative paradigm that proposes candidates directly through generative models. By reformulating the optimization of the acquisition function under a trust region as a problem of posterior inference, I will detail two of my recent frameworks—DiBO and CiBO. These works demonstrate how generative approaches can successfully and efficiently navigate high-dimensional spaces to solve complex problems in both unconstrained and constrained environments.

May 28th 3:00pm CET - Pieter Gijsbers

Title: OpenML: Insights from 10 years and more than a thousand papers

Abstract: TBA