BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//wp-events-plugin.com//7.2.3.1//EN
TZID:America/New_York
X-WR-TIMEZONE:America/New_York
BEGIN:VEVENT
UID:596@fds.yale.edu
DTSTART;TZID=America/New_York:20221219T123000
DTEND;TZID=America/New_York:20221219T133000
DTSTAMP:20250916T142120Z
URL:https://fds.yale.edu/events/fds-seminar-alkis-kalavasis-national-techn
 ical-university-of-athens/
SUMMARY:FDS Seminar: Alkis Kalavasis (National Technical University of Athe
 ns)
DESCRIPTION:"Efficient Algorithms and Computational Barriers in Reliable Ma
 chine Learning"Speaker: Alkis KalavasisNational Technical University of At
 hens\nAbstract: In this talk\, we will discuss the computational challenge
 s arising in various problems in Reliable Machine Learning. Reliable ML ai
 ms at the design of computationally efficient algorithms that provide guar
 antees such as robustness to biased data\, reproducibility and privacy. We
  firstly focus on the design of algorithms robust to biased and corrupted 
 observations. We begin with the problem of learning from coarse data. The 
 motivation behind this problem is that in many learning tasks one may not 
 have access to fine grained label information\; e.g.\, an image can be lab
 eled as husky\, dog\, or even animal depending on the expertise of the ann
 otator. We formalize these settings from the viewpoint of computational le
 arning theory and provide efficient algorithms and computational hardness 
 results. We then continue with the task of learning noisy linear label ran
 kings. Label ranking is the supervised task of learning a sorting function
  that maps feature vectors to rankings over a finite set of labels. We pro
 vide the first efficient algorithms for learning linear sorting functions 
 in the presence of bounded noise (an extension of the Massart noise condit
 ion to label rankings) under Gaussian marginals. Next\, we consider questi
 ons regarding responsibility aspects of ML systems. We study the important
  problem of reproducibility as an algorithmic property in decision-making 
 settings. We introduce the notion of reproducible policies in the context 
 of stochastic bandits\, one of the canonical problems in interactive learn
 ing. A policy in the bandit environment is called reproducible if it pulls
 \, with high probability\, the exact same sequence of arms in two differen
 t and independent executions (under independent reward realizations and sh
 ared internal randomness). We show that not only do reproducible policies 
 exist\, but also they achieve almost the same optimal (non-reproducible) r
 egret bounds in terms of the time horizon. In the end of the talk\, we wil
 l shortly discuss some on-going work on the complexity of min-max optimiza
 tion\, a fundamental problem in the area of equilibrium computation in mul
 ti-agent environments.
CATEGORIES:FDS Events,Postdoctoral Applicants
END:VEVENT
BEGIN:VTIMEZONE
TZID:America/New_York
X-LIC-LOCATION:America/New_York
BEGIN:STANDARD
DTSTART:20221106T010000
TZOFFSETFROM:-0400
TZOFFSETTO:-0500
TZNAME:EST
END:STANDARD
END:VTIMEZONE
END:VCALENDAR