BEGIN:VCALENDAR

VERSION:2.0

PRODID:-//wp-events-plugin.com//6.3//EN

TZID:Asia/Jerusalem

X-WR-TIMEZONE:Asia/Jerusalem
BEGIN:VEVENT

UID:273@dds.technion.ac.il

DTSTART;TZID=Asia/Jerusalem:20221208T101500

DTEND;TZID=Asia/Jerusalem:20221208T111500

DTSTAMP:20221201T115433Z

URL:https://dds.technion.ac.il/iemevents/hyper-networks-for-out-of-distrib
 ution-generalization-in-natural-language-processing/

SUMMARY:Hyper Networks for Out of Distribution Generalization in Natural La
 nguage Processing [	\n					Graduate Student Seminar\n					Seminars\n			\n	
 ]
DESCRIPTION:By: M.Sc. Tomer Volk\n Advisors: Roi Reichart\n Where: Bloomfie
 ld 526 From:\nTechnion\nAbstract:\n\nWhile Natural Language Processing (NL
 P) algorithms keep reaching unprecedented milestones\, out-of-distribution
  generalization is still challenging. In this talk we discuss the problem 
 of multi-source adaptation to unknown domains: Given labeled data from mul
 tiple source domains\, we aim to generalize to data drawn from target doma
 ins that are unknown to the algorithm at training time. We present an algo
 rithmic framework based on example-based Hypernetwork adaptation: Given an
  input example\, a T5 encoder-decoder first generates a unique signature w
 hich embeds this example in the semantic space of the source domains\, and
  this signature is then fed into a Hypernetwork which generates the weight
 s of the task classifier. In an advanced version of our model\, the learne
 d signature also serves for improving the representation of the input exam
 ple. In experiments with two tasks\, sentiment classification and natural 
 language inference\, across 29 adaptation settings\, our algorithms substa
 ntially outperform existing algorithms for this adaptation setup.
CATEGORIES:Graduate Student Seminar,Seminars
LOCATION:Bloomfield 526

END:VEVENT

BEGIN:VTIMEZONE

TZID:Asia/Jerusalem

X-LIC-LOCATION:Asia/Jerusalem

BEGIN:STANDARD

DTSTART:20221030T010000

TZOFFSETFROM:+0300

TZOFFSETTO:+0200

TZNAME:IST

END:STANDARD

END:VTIMEZONE
END:VCALENDAR