BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CREST - ECPv5.1.3//NONSGML v1.0//EN
CALSCALE:GREGORIAN
METHOD:PUBLISH
X-WR-CALNAME:CREST
X-ORIGINAL-URL:https://crest.science
X-WR-CALDESC:Events for CREST
BEGIN:VTIMEZONE
TZID:Europe/Helsinki
BEGIN:DAYLIGHT
TZOFFSETFROM:+0200
TZOFFSETTO:+0300
TZNAME:EEST
DTSTART:20230326T010000
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0300
TZOFFSETTO:+0200
TZNAME:EET
DTSTART:20231029T010000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTART;TZID=Europe/Helsinki:20230612T140000
DTEND;TZID=Europe/Helsinki:20230612T151500
DTSTAMP:20260711T063035
CREATED:20230607T105618Z
LAST-MODIFIED:20230607T105618Z
UID:15088-1686578400-1686582900@crest.science
SUMMARY:Nikita Zhivotovskiy (UC Berkeley) - “Optimal PAC Bounds without Uniform Convergence”
DESCRIPTION:Statistical Seminar: Every Monday at 2:00 pm.\nTime: 2:00 pm – 3:15 pm\nDate: 12th of June 2023\nPlace: Room 3001 \n  \nNikita Zhivotovskiy (UC Berkeley) – “Optimal PAC Bounds without Uniform Convergence” \n  \nAbstract: \nIn statistical learning theory\, the problem of determining sample complexity of realizable binary classification for VC classes was a longstanding challenge. Notable advancements by Simon and Hanneke established sharp upper bounds\, but their argument’s reliance on the uniform convergence principle curtailed its broader applicability to learning settings like multiclass classification. In this presentation\, we will discuss a new technique to resolve this limitation and introduce optimal high probability risk bounds within a framework that surpasses uniform convergence constraints. Beyond binary classification\, we will also delve into applications in scenarios where uniform convergence is notably sub-optimal. For multiclass classification\, we will prove an optimal risk bound that scales with the one-inclusion hypergraph density of the class\, effectively addressing the sub-optimality in the analysis by Daniely and Shalev-Shwartz. Additionally\, for realizable bounded regression with absolute loss\, we will derive an optimal risk bound based on a revised version of the scale-sensitive dimension\, thus refining the results of Bartlett and Long. This talk is based on the joint work with Ishaq Aden-Ali\, Yeshwanth Cherapanamjeri\, and Abhishek Shetty. \n  \n  \nOrganizers:\nCristina BUTUCEA (CREST)\, Alexandre TSYBAKOV (CREST)\, Karim LOUNICI (CMAP) \, Jaouad MOURTADA (CREST)\nSponsors:\nCREST-CMAP \n
URL:https://crest.science/event/nikita-zhivotovskiy-uc-berkeley-optimal-pac-bounds-without-uniform-convergence/
CATEGORIES:Statistics
ATTACH;FMTTYPE=:
END:VEVENT
END:VCALENDAR