BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Chicago
X-LIC-LOCATION:America/Chicago
BEGIN:DAYLIGHT
TZOFFSETFROM:-0600
TZOFFSETTO:-0500
TZNAME:CDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0500
TZOFFSETTO:-0600
TZNAME:CST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260202T201257Z
LOCATION:Second Floor Atrium
DTSTART;TZID=America/Chicago:20251121T080000
DTEND;TZID=America/Chicago:20251121T120000
UID:submissions.supercomputing.org_SC25_sess620_post259@linklings.com
SUMMARY:Intelligent Surrogates Pay Attention to Data, Improving Multi-Obje
 ctive HPC Optimization
DESCRIPTION:Ashna Nawar Ahmed (Texas State University, Oak Ridge National 
 Laboratory (ORNL)); Banooqa Banday (Texas State University); Terry Jones (
 Oak Ridge National Laboratory (ORNL)); and Tanzima Z. Islam (Texas State U
 niversity)\n\nHigh performance computing (HPC) schedulers must balance run
 time and power. We present a surrogate-assisted multi-objective Bayesian o
 ptimization (MOBO) framework using TabNet regressors and models trained on
  attention-based embeddings, coupled with active-learning sample selection
 . The surrogates predict runtime and power, enabling MOBO to efficiently d
 iscover Pareto-optimal node allocations. We quantify trade-offs with Paret
 o fronts, hypervolume (HV), and Spread across PM100 and Adastra production
  traces. MOBO improves HV over single-objective baselines by 24% (PM100) a
 nd 37% (Adastra) and attains lower Spread in 75% of surrogate families. Ac
 tive learning reduces evaluations by ~53%–70%. To our knowledge, this is t
 he first demonstration of embedding-informed surrogates for MOBO applied t
 o HPC job scheduling traces, optimizing runtime–power trade-offs on produc
 tion datasets.\n\nTag: Research & ACM SRC Posters\n\nRegistration Category
 : Technical Program Reg Pass\n\nSession Chairs: Kento Sato (RIKEN Center f
 or Computational Science (R-CCS)); Anja Gerbes (Georg-August-Universität G
 öttingen); and Chris Schlipalius (Pawsey Supercomputing Research Centre; C
 ommonwealth Scientific and Industrial Research Organisation (CSIRO), Austr
 alia)\n\n
END:VEVENT
END:VCALENDAR