BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:IC Colloquium: Generative AI Models that Learn from Bad Data
DTSTART:20260209T101500
DTEND:20260209T111500
DTSTAMP:20260916T063924Z
UID:0e6fb150a15ae40e3150f796f96dd6baf5ba713bcf5bfc4febd7d6f0
CATEGORIES:Conferences - Seminars
DESCRIPTION:By: Giannis Daras - MIT\nIC Faculty candidate\n\nAbstract\nIn 
 this talk\, we will introduce a principled and practical framework for tra
 ining generative models with imperfect samples. Recent progress in Generat
 ive AI is fuelled by the availability of large-scale\, high-quality datas
 ets. However\, in many practical applications\, high-quality samples are s
 carce\, expensive\, or altogether impossible to obtain. Even in data-rich 
 settings (such as the image domain)\, we are approaching the limits of hig
 h-quality human-generated data. We will show how to leverage imperfect dat
 a sources\, including low-quality\, corrupted\, synthetic\, and out-of-dis
 tribution samples\, which are cheaper and more widely available. We will i
 nstantiate the framework for diffusion models\, one of the most powerful c
 lasses of generative models\, and highlight applications in Computer Visio
 n and Computational Biology that achieve state-of-the-art results for imag
 e generation and de novo protein design\, respectively. Finally\, we will 
 discuss extensions to the textual domain\, implications for memorization\,
  privacy\, data pricing and data collection\, and a future where generativ
 e models and datasets co-evolve\, refining each other to transcend the ori
 ginal data and explore the solution space.\n\nBio\nGiannis Daras is a Post
 doctoral Associate at the Massachusetts Institute of Technology (MIT) supe
 rvised by Antonio Torralba and Costis Daskalakis. Giannis obtained his Ph.
 D. from the Computer Science department of UT Austin under the supervision
  of Alex Dimakis. Giannis works on important practical and theoretical que
 stions around deep generative models with a focus on training and sampling
  generative models in the presence of data corruption. His work has found 
 applications across scientific fields such as Computer Vision\, Computatio
 nal Biology\, Medical Imaging\, Robotics\, Economics\, and Neuroscience.\n
 \nGiannis has been nominated as a Rising Star in AI by the University of M
 ichigan\, has earned the Best Contribution Award at the Biomedical and Ast
 ronomical Signal Processing (BASP) conference\, and has published 23 resea
 rch papers (including 13 first-author works) at top-tier Machine Learning 
 venues\, including one Oral Presentation (top 0.3%) and two Spotlight pres
 entations (top 3.2%) at NeurIPS. Giannis has also been supported by multip
 le fellowships\, including the Graduate Dean’s Prestigious Fellowship (U
 T Austin)\, Onassis\, Bodossakis\, Leventis and Gerondellis Ph.D. Fellowsh
 ips. His work has sparked the interest of the public following coverage by
  news outlets such as The Independent and the New York Post.\n\nMore infor
 mation
LOCATION:BC 420 https://plan.epfl.ch/?room==BC%20420 https://epfl.zoom.us/
 j/66602224057
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR