BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Los_Angeles
X-LIC-LOCATION:America/Los_Angeles
BEGIN:DAYLIGHT
TZOFFSETFROM:-0800
TZOFFSETTO:-0700
TZNAME:PDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0700
TZOFFSETTO:-0800
TZNAME:PST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260402T024533Z
LOCATION:3004\, Level 3
DTSTART;TZID=America/Los_Angeles:20250623T171500
DTEND;TZID=America/Los_Angeles:20250623T173000
UID:dac_DAC 2025_sess125_RESEARCH1225@linklings.com
SUMMARY:ArbiterQ: Improving QNN Convergency and Accuracy by Applying Perso
 nalized Model on Heterogeneous Quantum Devices
DESCRIPTION:Tianyao Chu, Siwei Tan, and Liqiang Lu (Zhejiang University); 
 Jingwen Leng and Fangxin Liu (Shanghai Jiao Tong University); and Conglian
 g Lang, Yifan Guo, and Jianwei Yin (Zhejiang University)\n\nIn the current
  NISQ era, the performance of QNN models is strictly hindered by the limit
 ed qubit number and inevitable noise. A natural idea to improve the robust
 ness of QNN is to involve multiple quantum devices. Nevertheless, due to t
 he heterogeneity and instability of quantum devices (e.g., noise, frequent
  online/offline), training and inference on distributed quantum devices ma
 y even destroy the accuracy.\n\nIn this paper, we propose ArbiterQ, a comp
 rehensive QNN framework designed for efficient and high-accuracy training 
 and inference on heterogeneous QPUs. The main innovation of ArbiterQ is it
  applies personalized models for each QPU via two uniform QNN representati
 ons: model vector and behavioral vector. The model vector specifies the lo
 gical-level parameters in the QNN model, while the behavioral vector captu
 res the hardware-level features when implementing the QNN circuit. In this
  manner, by sharing the gradient among QPUs with similar behavioral vector
 s, we can effectively leverage parallelism while considering heterogeneity
 . We also propose shot-oriented inference scheduling, which is a much more
  fine-grained scheduling that can improve accuracy and balance the workloa
 d. The experiments show that ArbiterQ accelerates the training process by 
 4.03X with 7.87% loss reduction, compared with the previous distributed QN
 N framework EQC.\n\nTopics: Design\n\nTracks: DES6: Quantum Computing\n\nS
 ession Chairs: Jinglei Cheng (University of Pittsburgh) and Himanshu Thapl
 iyal (University of Tennessee)\n\n
END:VEVENT
END:VCALENDAR