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:20260402T024534Z
LOCATION:3004\, Level 3
DTSTART;TZID=America/Los_Angeles:20250623T141500
DTEND;TZID=America/Los_Angeles:20250623T143000
UID:dac_DAC 2025_sess145_RESEARCH993@linklings.com
SUMMARY:Generalizable Lithographic Hotspot Detection Using Asynchronous Me
 ta-Learning with Only One Shot
DESCRIPTION:Cong Jiang, Yujia Wang, and Dan Feng (Huazhong University of S
 cience and Technology); Haoyu Yang (Nvidia); and Kang Liu (Huazhong Univer
 sity of Science and Technology)\n\nWith integrated circuits shrinking in f
 eature size, layout printability has become increasingly challenging, maki
 ng lithographic hotspot detection ever-crucial in computer-aided design (C
 AD) flows. In recent years, numerous studies have explored deep learning t
 o detect lithographic hotspots, offering promising results. However, neura
 l networks can easily be biased and overfit when lacking sufficient traini
 ng data, especially in the CAD domain. A generalizable DL-based hotspot de
 tector should learn the genuine lithography principle and ensure consisten
 t accuracy across layouts from various designs at the same technology node
 , regardless of their varying design styles. However, we find that existin
 g convolutional neural network (CNN)-based hotspot detectors fail to gener
 alize to different circuit layouts other than the design it has been train
 ed for. To this end, we propose a few-shot learning-based framework for ge
 neralizable CNN-based hotspot detection. We develop a meta-learning scheme
  that asynchronously updates the CNN feature extraction and classification
  component to obtain a meta-initialized model that can quickly adapt to ne
 w designs using as few as one training layout clip. We propose a layout to
 pology-based sampling strategy for few-shot adaptation to enhance generali
 zation stability. Experimental results on ICCAD 2012 and 2019 datasets sho
 w that our framework enables superior generalization capabilities than pri
 or arts on unseen new designs.\n\nTopics: EDA\n\nTracks: EDA8: Design for 
 Manufacturing and Reliability\n\nSession Chairs: Iris Hui-Ru Jiang (Nation
 al Taiwan University) and Luigi Capodieci (Synopsys)\n\n
END:VEVENT
END:VCALENDAR