ccanonne.github.io

FOCS 2023 Workshop: Algorithmic Aspects of High-Dimensional Probabilistic Models

Organized by Arnab Bhattacharyya (National University of Singapore) and Cl矇ment Canonne (University of Sydney)

Dates: Nov 67, 2023

High-dimensional probability distributions are currently omnipresent as models for complex data. Probabilistic models are used in an amazingly diverse array of data analysis tasks, from understanding sociological behavior to enabling visual object recognition, speech perception, and natural language comprehension. High-dimensional probabilistic models take various forms: classically-studied models such as multivariate Gaussians and Erds-R矇nyi graphs, models with roots in statistical physics such as stochastic block models and Ising models, probabilistic graphical models such as Bayesian networks and Markov random fields, as well as the class of implicit generative models, such as generative adversarial networks and large language models, that have taken the world by storm recently.

The goal of the proposed workshop is to highlight current trends in research on algorithmic aspects of high-dimensional probabilistic models. Specifically, we will focus on recent works that identify new directions for the TCS community, so as to encourage newcomers to delve into the important problems in the area. The workshop is particularly timely, as it is increasingly in vogue to be presented with generative models that sample from a distribution D, with the goal being to efficiently learn information about D. What are rigorous guarantees that can be made when D is high-dimensional and complex? Better understanding of the computational aspects of such questions can be potentially impactful for a number of areas, e.g., security, machine learning, probabilistic programming, etc.

Program: All talks will take place in room Sequoia D, 910:15am.

The program will feature talks on recent and exciting results in this area, and will gather open problems in the topic from and for the participants:

Call for Open Problems: To suggest an open problem, please fill this Google form

Dinner: to gauge interest in a dinner on November 6 (first day) for the workshop participants, please fill this other Google form