BEGIN:VCALENDAR
CALSCALE:GREGORIAN
VERSION:2.0
METHOD:PUBLISH
PRODID:-//Drupal iCal API//EN
X-WR-TIMEZONE:America/New_York
BEGIN:VTIMEZONE
TZID:America/New_York
BEGIN:DAYLIGHT
TZOFFSETFROM:-0500
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
DTSTART:20070311T020000
TZNAME:EDT
TZOFFSETTO:-0400
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0400
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
DTSTART:20071104T020000
TZNAME:EST
TZOFFSETTO:-0500
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
SEQUENCE:1
X-APPLE-TRAVEL-ADVISORY-BEHAVIOR:AUTOMATIC
UID:242756
DTSTAMP:20260915T104552Z
DTSTART;TZID=America/New_York:20260918T100000
DTEND;TZID=America/New_York:20260918T110000
URL;TYPE=URI:https://www.wpi.edu/news/calendar/events/robotics-engineering-
 phd-speaking-and-writing-qualifiers-presentation-fangzhou-lin
SUMMARY:Robotics Engineering PhD Speaking and Writing Qualifiers Presentati
 on: Fangzhou Lin
DESCRIPTION:NexusFlow: Unifying Disparate Tasks under Partial Supervision v
 ia Invertible Flow Networks\n\n\n\n      \n      \n\n\n\nAbstract: Partial
 ly Supervised Multi-Task Learning (PS-MTL) aims to leverage knowledge acro
 ss tasks when annotations are incomplete. Existing approaches\, however\, 
 have largely focused on the simpler setting of homogeneous\, dense predict
 ion tasks\, leaving the more realistic challenge of learning from structur
 ally diverse tasks unexplored. To this end\, we introduce NexusFlow\, a no
 vel\, lightweight\, and plug-and-play framework effective in both settings
 . NexusFlow introduces a set of surrogate networks with invertible couplin
 g layers to align the latent feature distributions of tasks\, creating a u
 nified representation that enables effective knowledge transfer. The coupl
 ing layers are bijective\, preserving information while mapping features i
 nto a shared canonical space. This invertibility avoids representational c
 ollapse and enables alignment across structurally different tasks without 
 reducing expressive capacity. We first evaluate NexusFlow on the core chal
 lenge of domain-partitioned autonomous driving\, where dense map reconstru
 ction and sparse multi-object tracking are supervised in different geograp
 hic regions\, creating both structural disparity and a strong domain gap. 
 NexusFlow sets a new state-of-the-art result on nuScenes\, outperforming s
 trong partially supervised baselines. To demonstrate generality\, we furth
 er test NexusFlow on NYUv2 using three homogeneous dense prediction tasks\
 , segmentation\, depth\, and surface normals\, as a representative N-task 
 PS-MTL scenario. NexusFlow yields consistent gains across all tasks\, conf
 irming its broad applicability.\nAdvisor: Professor Haichong ZhangCommitte
 e: Professor Guanrui Li\, Professor Ziming Zhang\nZoom link: https://wpi.z
 oom.us/j/8204769623
END:VEVENT
END:VCALENDAR