BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Europe/Stockholm
X-LIC-LOCATION:Europe/Stockholm
BEGIN:DAYLIGHT
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20250822T115805Z
LOCATION:Campussaal - Plenary Room
DTSTART;TZID=Europe/Stockholm:20250617T103000
DTEND;TZID=Europe/Stockholm:20250617T110000
UID:submissions.pasc-conference.org_PASC25_sess150_pos125@linklings.com
SUMMARY:P40 - Spectral Methods for the Clustering of Cyclic and Acyclic Gr
 aphs
DESCRIPTION:Jacopo Palumbo (Università della Svizzera italiana, Politecnic
 o di Milano); Dimosthenis Pasadakis (Università della Svizzera italiana); 
 Albert-Jan Yzelman (Huawei); and Olaf Schenk (Università della Svizzera it
 aliana)\n\nTraditional spectral clustering methods are designed for undire
 cted graphs and fail to capture the directionality of the edges and of the
  connections between the clusters. The aim of our work is centered around 
 developing novel spectral methods for the spectral clustering of directed 
 graphs with block-cyclic and block-acyclic structures. Block-cyclic instan
 ces are obtained from phenomena with recurrent patterns, while block-acycl
 ic ones capture hierarchical relationships, and usually appear in real-wor
 ld scenarios such as task scheduling between processors and trophic networ
 ks. We extend previously introduced spectral methods for the clustering of
  block-cyclic and block-acyclic graphs to novel algorithms, employing nonl
 inear graph Laplacians, that provide sharper approximations for the direct
 ed graph cuts, resulting in higher clustering accuracy. Additionally, we l
 everage diffusion principles in the transition matrices under question, to
  effectively minimize the normalized cut between the partitions. The effec
 tiveness of the introduced algorithms is validated through a series of exp
 eriments on synthetic and real-world graphs. The performance of the algori
 thms is measured both with metrics based on the quality of the graph-cut, 
 and with metrics based on the accuracy of labels retrieval.\n\nSession Cha
 ir: David Moxey (King's College London)\n\n
END:VEVENT
END:VCALENDAR