BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Robust Inference and Local Algorithms
DTSTART:20170615T160000
DTEND:20170615T164500
DTSTAMP:20260407T162258Z
UID:59b73130f95291c2590618682b89f146f32ef6963b7200f2656698e8
CATEGORIES:Conferences - Seminars
DESCRIPTION:Yishay Mansour\, Tel Aviv University\nRobust inference is an e
 xtension of probabilistic inference\, where some of the observations may b
 e adversarially corrupted. We limit the adversarial corruption to a finite
  set of modification rules. We model robust inference as a zero-sum game b
 etween an adversary\, who selects a modification rule\, and a predictor\, 
 who wants to accurately predict the state of nature.\nThere are two varian
 ts of the model\, one where the adversary needs to pick the modification r
 ule in advance and one where the adversary can select the modification rul
 e after observing the realized uncorrupted input. For both settings we der
 ive efficient near optimal policy runs in polynomial time. Our efficient a
 lgorithms are based on methodologies for developing local computation algo
 rithms.\nWe also consider a learning setting where the predictor receives 
 a set of uncorrupted inputs and their classification. The predictor needs 
 to select a hypothesis\, from a known set of hypotheses\, and is tested on
  inputs which the adversary corrupts. We show how to utilize an ERM oracle
  to derive a near optimal predictor strategy\, namely\, picking a hypothes
 is that minimizes the error on the corrupted test inputs.\nBased on joint 
 works with Uriel Feige\, Aviad Rubinstein\, Robert Schapire\, Moshe Tennen
 holtz\, Shai Vardi.
LOCATION:BC 420 https://plan.epfl.ch/?room==BC%20420
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR