BEGIN:VCALENDAR
VERSION:2.0
PRODID:icalendar-ruby
CALSCALE:GREGORIAN
X-WR-CALNAME:Are All Features Created Equal?
X-WR-TIMEZONE:Eastern Time (US & Canada)
BEGIN:VEVENT
DTSTAMP:20260517T032215Z
UID:tag:localist.com\,2008:EventInstance_33077030345933
DTSTART:20200331T203000Z
DTEND:20200331T213000Z
DESCRIPTION:LIDS Seminar Series\n\nSpeaker: Aleksander Mądry\n\nAffiliatio
 n: MIT\n\nAbstract\n\nOur machine learning models have attained impressive
  accuracy on many benchmark tasks. Yet\, these models remain remarkably br
 ittle---small perturbations of natural inputs can completely degrade their
  performance.\n\nWhy is this the case?\n\nIn this talk\, we show that this
  brittleness can\, in part\, be attributed to the fact that our models oft
 en make decisions in a very different way than humans do. Viewing neural n
 etworks as feature extractors\, we study how features extracted by neural 
 networks may diverge from those used by humans\, and how adversarially rob
 ust models seem to make progress towards bridging this gap.\n\nBiography \
 n\nAleksander Madry is a Professor of Computer Science in the MIT EECS Dep
 artment and a Principal Investigator in the MIT Computer Science and Artif
 icial Intelligence Laboratory (CSAIL). He received his PhD from MIT in 201
 1 and\, prior to joining the MIT faculty\, he spent some time at Microsoft
  Research New England and on the faculty of EPFL.\n\nAleksander's research
  interests span algorithms\, continuous optimization\, the science of deep
  learning\, and understanding machine learning from a robustness perspecti
 ve. His work has been recognized with a number of awards\, including an NS
 F CAREER Award\, an Alfred P. Sloan Research Fellowship\, an ACM Doctoral 
 Dissertation Award Honorable Mention\, and Presburger Award.\n\nJoin Zoom 
 Meeting   https://mit.zoom.us/j/268033196\n\nMeeting ID   268 033 196\n\nJ
 oin by SIP   268033196@zoomcrc.com\n\nJoin by Skype for Business   https:/
 /mit.zoom.us/skype/268033196\n\n____________________________________\n\nTh
 e LIDS Seminar Series features distinguished speakers who provide an overv
 iew of a research area\, as well as exciting recent progress in that area.
  Intended for a broad audience\, seminar topics span the areas of communic
 ations\, computation\, control\, learning\, networks\, probability and sta
 tistics\, optimization\, and signal processing.
GEO:42.361965;-71.090261
LOCATION:Building 32\, 141
SUMMARY:Are All Features Created Equal?
URL;VALUE=URI:https://calendar.mit.edu/event/are_all_features_created_equal
CATEGORIES:Conferences/Seminars/Lectures
END:VEVENT
END:VCALENDAR