BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Talk of Professor Pradeep Ravikumar\, School of Computer Science a
 t Carnegie Mellon University
DTSTART:20220914T111500
DTEND:20220914T121500
DTSTAMP:20260427T230602Z
UID:2ebf8c8074785053844d50110d6014278e8c9c4618bfa5052f22595d
CATEGORIES:Conferences - Seminars
DESCRIPTION:Professor Pradeep Ravikumar\, School of Computer Science at Ca
 rnegie Mellon University\nTitle: Building Robust Ensembles via Margin Boos
 ting\n\nAbstract: In the context of adversarial robustness\, a single mode
 l does not usually have enough power to defend against all possible advers
 arial attacks\, and as a result\, has sub-optimal robustness. Consequently
 \, an emerging line of work has focused on learning an ensemble of neural 
 networks to defend against adversarial attacks. In this work\, we take a p
 rincipled approach towards building robust ensembles. We view this problem
  from the perspective of solving a margin-boosting game\, and develop an a
 lgorithm for learning an ensemble with maximum margin. Through extensive e
 mpirical evaluation on benchmark datasets\, we show that our algorithm not
  only outperforms existing ensembling techniques\, but also large models t
 rained in an end-to-end fashion. An important byproduct of our work is a m
 argin-maximizing cross-entropy (MCE) loss\, which is a better alternative 
 to the standard cross-entropy (CE) loss. Empirically\, we show that replac
 ing the CE loss in state-of-the-art adversarial training techniques with o
 ur MCE loss leads to significant performance improvement.\nJoint work with
  Arun Sai Suggala\, Dinghuai Zhang\, Hongyang Zhang\, Aaron Courville\, Yo
 shua Bengio\n\nBio: Pradeep Ravikumar is a Professor in the Machine Learni
 ng Department\, School of Computer Science at Carnegie Mellon University. 
 He was previously an Associate Director at the Center for Big Data Analyti
 cs\, at the University of Texas at Austin. His thesis has received honorab
 le mentions in the ACM SIGKDD Dissertation award and the CMU School of Com
 puter Science Distinguished Dissertation award. He is a Sloan Fellow\, a S
 iebel Scholar\, a recipient of the NSF CAREER Award\, and was Program Chai
 r for the International Conference on Artificial Intelligence and Statisti
 cs (AISTATS) in 2013. He is Associate Editor-in-Chief for IEEE Transaction
 s on Pattern Analysis and Machine Intelligence (TPAMI)\, and action editor
  for the Machine Learning journal\, and the Journal of Machine Learning Re
 search.\n\nDr. Ravikumar's research group at CMU works on the foundations 
 of statistical machine learning\, with recent focus on "next generation" m
 achine learning systems\, that are explainable\, robust to train and test 
 time corruptions\, and resilient to distribution shifts\, and are learnt u
 nder resource constraints by leveraging or discovering various notions of 
 "structure" and domain knowledge.
LOCATION:BM 5202 https://plan.epfl.ch/?room==BM%205202 https://epfl.zoom.u
 s/j/81786690250
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR