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:Level 2 Lobby
DTSTART;TZID=America/Los_Angeles:20250623T180000
DTEND;TZID=America/Los_Angeles:20250623T190000
UID:dac_DAC 2025_sess287_LBR105@linklings.com
SUMMARY:Late Breaking Results: Customized Diffusion Model Empowered by Het
 erogeneous Graph Network for Effective Floorplanning
DESCRIPTION:Xinglin Zheng, Hao Gu, Keyu Peng, and Youwen Wang (Southeast U
 niversity); Wenxing Zhu (Fuzhou University); and Ziran Zhu (Southeast Univ
 ersity)\n\nIn this paper, we propose a customized diffusion model to direc
 tly generate high-quality initial floorplans. \nBy leveraging a classical 
 analytical-based floorplanner on top of this initial floorplan, the final 
 floorplanning results are significantly improved.\nTo enhance feature extr
 action, a heterogeneous graph neural network (HGNN) is developed to explic
 itly incorporate block-to-block and pin-to-block relationships from the ne
 tlist during the diffusion process.\nAdditionally, a novel guidance sampli
 ng function is introduced to optimize both wirelength and overlap, effecti
 vely reducing the required sampling steps while maintaining competitive in
 itial solutions.\nExperimental results demonstrate that integrating our pr
 oposed diffusion model with an advanced analytical-based floorplanner achi
 eves at least 4.8\% reduction in runtime and 3.0\% reduction in HPWL compa
 red to the original floorplanner and other diffusion-based methods.\n\n
END:VEVENT
END:VCALENDAR