Vladimir Pestov: Some set-theoretic motives in statistical learning theory II

(reminder: no seminar tomorrow)

(announcement: the VU University Amsterdam is firing its
topology/geometry faculty—please sign the petition to save the
program! http://www.gopetition.com/petitions/save-pure-mathematics-at-the-vu-university-amsterdam.html)

Toronto Set Theory Seminar
Friday, May 6 from 1:30 to 3pm
Fields Institute, Room 210

Speaker: Vladimir Pestov (University of Ottawa)

Title: Some set-theoretic motives in statistical learning theory II

Abstract: We will discuss a few instances where the Vapnik-Chervonenkis theory of statistical learning has potentially interesting overlaps with set theory / model theory / extremal combinatorics. This will be illustrated on speaker’s recent results about PAC learnability over non-atomic measures (a solution of a problem by Vidyasagar with the help of Martin’s Axiom), as well as learnability over exchangeable data inputs, and a reputed 25-year old open problem concerning the so-called sample compression schemes.

Attachments

One response to “Vladimir Pestov: Some set-theoretic motives in statistical learning theory II

Leave a Reply

Your email address will not be published. Required fields are marked *

Time limit is exhausted. Please reload CAPTCHA.