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:20241120T082409Z
LOCATION:HG F 3
DTSTART;TZID=Europe/Stockholm:20240604T173000
DTEND;TZID=Europe/Stockholm:20240604T180000
UID:submissions.pasc-conference.org_PASC24_sess112_msa261@linklings.com
SUMMARY:Learning to Predict and Improve Build Successes in Package Ecosyst
 ems Using Graph Neural Networks
DESCRIPTION:Minisymposium\n\nHarshitha Menon (Lawrence Livermore National 
 Laboratory)\n\nModern software has reached an unprecedented level of compl
 exity, consisting of tens or even hundreds of dependencies on various pack
 ages. In order to tackle this complexity, software ecosystems rely on auto
 mated package managers to analyze compatibility constraints among differen
 t packages and select a compatible set of package versions to install. Cur
 rent approaches rely on experts with in-depth knowledge of packages and co
 nstraints to identify compatible versions. In practice, users often have t
 o explore different choices of package versions to find an appropriate one
  that builds successfully. In this talk, we present a tool, called BuildCh
 eck, to understand build incompatibilities, predict bad configurations, an
 d assist developers in managing version constraints. We combine the capabi
 lities of Graph Neural Networks and advanced package management technologi
 es to offer solutions for managing package dependencies. Our tool, BuildCh
 eck, evaluated on E4S software ecosystem consisting of 45, 837 data points
  can predict build outcomes with 91% accuracy eliminating very expensive t
 rial-and-error exercises to find working builds. Furthermore, our novel se
 lf-supervised pre-training method using masked modeling was shown to impro
 ve the prediction accuracy when only a limited amount of data is available
 .\n\nDomain: Computational Methods and Applied Mathematics\n\nSession Chai
 rs: Florina Ciorba (University of Basel) and Harshitha Menon (Lawrence Liv
 ermore National Laboratory)
END:VEVENT
END:VCALENDAR