BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//132.216.98.100//NONSGML kigkonsult.se iCalcreator 2.20.4//
BEGIN:VEVENT
UID:20251002T053434EDT-8301WvvTUS@132.216.98.100
DTSTAMP:20251002T093434Z
DESCRIPTION:\n	Convergence Guarantees for Adversarially Robust Classifiers\n
 \n	 \n\n\nAbstract:\n\nNeural networks can be trained to classify images an
 d achieve high levels of accuracy. However\, researchers have discovered t
 hat well-targeted perturbations of an image can completely fool a trained 
 classifier\, even in cases where the modified image is visually indistingu
 ishable from the original. This has sparked many new approaches to classif
 ication which include an adversary in the training process: such an advers
 ary can improve robustness and generalization properties at the cost of de
 creased accuracy and increased training time. In this presentation\, I wil
 l explore the connection between a certain class of adversarial training p
 roblems and the Bayes classification problem for binary classification. In
  particular\, robustness can be encouraged by adding a regularizing nonloc
 al perimeter term\, providing a strong connection to classical studies of 
 perimeter. Borrowing tools from geometric measure theory\, I will show the
  Hausdorff convergence of adversarially robust classifiers to Bayes classi
 fiers as the strength of adversary decreases to 0. In this way\, the theor
 etical results discussed in the presentation provide a rigorous comparison
  with the standard Bayes classification problem.\n\nSpeaker\n\nRachel Morr
 is received her PhD from NC State this summer under the supervision of Rya
 n Murray. Her PhD work studied data-driven optimization problems through t
 he lens of calculus of variations and geometric measure theory. Currently\
 , Rachel is a postdoc at Concordia University\, working with Jason Brambur
 ger and Simone Brugiapaglia\n
DTSTART:20251003T193000Z
DTEND:20251003T200000Z
LOCATION:Room 1104\, Burnside Hall\, CA\, QC\, Montreal\, H3A 0B9\, 805 rue
  Sherbrooke Ouest
SUMMARY:Rachel Morris (Concordia University)
URL:/science/channels/event/rachel-morris-concordia-un
 iversity-368046
END:VEVENT
END:VCALENDAR