BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//https://caida.ubc.ca//NONSGML iCalcreator 2.41.92//
CALSCALE:GREGORIAN
METHOD:PUBLISH
UID:31383962-6139-4461-a430-663832376633
X-WR-RELCALID:efc09d74-9c93-479e-a94f-485231ddccde
X-WR-TIMEZONE:America/Vancouver
X-WR-CALNAME:Efficient Inference of Mixture-of-Experts (MoE)-based Large Mo
 dels with Theoretical Guarantees - Meng Wang\, Professor\, Rensselaer Poly
 technic Institute
BEGIN:VTIMEZONE
TZID:America/Vancouver
TZUNTIL:20270314T100000Z
BEGIN:STANDARD
TZNAME:PST
DTSTART:20241103T020000
TZOFFSETFROM:-0700
TZOFFSETTO:-0800
RDATE:20251102T020000
RDATE:20261101T020000
END:STANDARD
BEGIN:DAYLIGHT
TZNAME:PDT
DTSTART:20250309T020000
TZOFFSETFROM:-0800
TZOFFSETTO:-0700
RDATE:20260308T020000
END:DAYLIGHT
END:VTIMEZONE
BEGIN:VEVENT
UID:b4a3e525-60c2-481c-82df-26e0b732310d
DTSTAMP:20260220T014414Z
CLASS:PUBLIC
CREATED:20250709T194331Z
DESCRIPTION:Abstract: Mixture-of-Experts (MoE) architectures have emerged a
 s a powerful paradigm for scaling large models by routing inputs to specia
 lized subnetworks (experts)\, achieving impressive performance with reduce
 d computation during training. However\, efficient inference of MoE models
  remains challenging due to memory and computational overhead\, especially
  when deployed in resource-constrained environments. In this talk\, I will
  first introduce a provably efficient expert pruning method for fine-tuned
  MoE models\, which preserves test-time accuracy by pruning experts with m
 inimal change in router…
DTSTART;TZID=America/Vancouver:20250715T140000
DTEND;TZID=America/Vancouver:20250715T150000
LAST-MODIFIED:20250709T195414Z
LOCATION:UBC Vancouver Campus\, Fried Kaiser (KAIS) building\, Room 2020/20
 30\, 2332 Main Mall
SUMMARY:Efficient Inference of Mixture-of-Experts (MoE)-based Large Models 
 with Theoretical Guarantees - Meng Wang\, Professor\, Rensselaer Polytechn
 ic Institute
TRANSP:OPAQUE
URL:https://caida.ubc.ca/event/efficient-inference-mixture-experts-moe-base
 d-large-models-theoretical-guarantees-meng-wang
END:VEVENT
END:VCALENDAR