BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Date iCal//NONSGML kigkonsult.se iCalcreator 2.20.2//
METHOD:PUBLISH
X-WR-CALNAME;VALUE=TEXT:Eventi DIAG
BEGIN:VTIMEZONE
TZID:Europe/Paris
BEGIN:STANDARD
DTSTART:20221030T030000
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
RDATE:20231029T030000
TZNAME:CET
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20230326T020000
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
END:DAYLIGHT
END:VTIMEZONE
BEGIN:VEVENT
UID:calendar.25701.field_data.0@www.u-gov-ricerca.uniroma1.it
DTSTAMP:20260407T200310Z
CREATED:20230508T174644Z
DESCRIPTION:Title: Multi-Swap Local Search for K-means. Abstract: The k-mea
 ns++ algorithm of Arthur and Vassilvitskii (SODA 2007) is often the practi
 tioners' choice algorithm for solving the k-means clustering problem and i
 s known to give an O(log k)-approximation in expectation.Lattanzi and Sohl
 er (ICML 2019) proposed augmenting k-means++ with O(k log log k) local sea
 rch steps to yield a O(1)-approximation to the k-means clustering problem\
 , where O(1) hides a large constant. Here we  generalise their algorithm a
 llowing it to swap multiple centers at the same time. Our algorithm achiev
 es a 9 + epsilon approximation ratio\, which is the best possible for loca
 l search. While the same approximation ratio can be achieved combining sev
 eral known techniques\, our algorithm is both asymptotically faster and mo
 re practical. 
DTSTART;TZID=Europe/Paris:20230511T150000
DTEND;TZID=Europe/Paris:20230511T150000
LAST-MODIFIED:20230508T180525Z
LOCATION:DIAG - Aula Magna
SUMMARY:Multi-Swap Local Search for K-means: Dr. Lorenzo Beretta (Univ. of 
 Copenhagen)  - Dr. Lorenzo Beretta (Univ. of Copenhagen)
URL;TYPE=URI:http://www.u-gov-ricerca.uniroma1.it/node/25701
END:VEVENT
END:VCALENDAR