BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Namur Institute For Complex Systems - ECPv6.15.20//NONSGML v1.0//EN
CALSCALE:GREGORIAN
METHOD:PUBLISH
X-ORIGINAL-URL:https://www.naxys.be
X-WR-CALDESC:Events for Namur Institute For Complex Systems
REFRESH-INTERVAL;VALUE=DURATION:PT1H
X-Robots-Tag:noindex
X-PUBLISHED-TTL:PT1H
BEGIN:VTIMEZONE
TZID:Europe/Brussels
BEGIN:DAYLIGHT
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
DTSTART:20230326T010000
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
DTSTART:20231029T010000
END:STANDARD
BEGIN:DAYLIGHT
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
DTSTART:20240331T010000
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
DTSTART:20241027T010000
END:STANDARD
BEGIN:DAYLIGHT
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
DTSTART:20250330T010000
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
DTSTART:20251026T010000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTART;TZID=Europe/Brussels:20241115T130000
DTEND;TZID=Europe/Brussels:20241115T140000
DTSTAMP:20260522T080327
CREATED:20241030T101730Z
LAST-MODIFIED:20241106T122335Z
UID:2224-1731675600-1731679200@www.naxys.be
SUMMARY:Serge Gratton (Toulouse INP)
DESCRIPTION:Title : Adaptive regularization algorithms for Objective-Function-Free Optimization \nAbstract :  \nThis work is about the fastest known optimization method in terms of complexity\, which does not require the evaluation of the objective function. Such methods\, termed Objective-Function-Free Optimization (OFFO)\, have gained popularity recently\, especially in noisy problem contexts\, including deep learning applications. We present an adaptive regularization OFFO algorithm for unconstrained nonconvex optimization. This algorithm is part of the adaptive regularization methods family\, which is known for its optimal worst-case complexity results in the conventional framework where the objective function is evaluated. We demonstrate that these outstanding complexity bounds are maintained with the new algorithm\, even though it relies on significantly less information. Specifically\, we show that using derivatives of degree one to p\, the algorithm will find an 𝛆1-approximate first-order minimizer in at most O(𝛆-(p+1)/p) iterations\, and an (𝛆1\, 𝛆2)-approximate second-order minimizer in at most O(𝛆-(p+1)/p\, 𝛆-(p+1)/(p-1)}) iterations. As a particular case\, the algorithm\, when using first and second derivatives and applied to functions with Lipschitz continuous Hessian\, will find an iterate xk where the gradient’s norm is less than 𝛆1 in at most O(𝛆-3/2) iterations. Numerical experiments demonstrate the excellent performance of this method for noisy problems\, highlighting its robustness and efficiency across various scenarios. \nThe seminar will take place in Room S08 at the Faculty of Sciences.
URL:https://www.naxys.be/event/serge-gratton-toulouse-inp/
CATEGORIES:NAXYS Seminar
END:VEVENT
END:VCALENDAR