BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:What to do When Data is Shifting? - Invariance Objectives\, Enviro
 nment Discovery and Expert Augmentations
DTSTART:20220531T110000
DTEND:20220531T120000
DTSTAMP:20260408T092625Z
UID:6cb98f24ab7274dea93d79b32c599867667f26243c4bc3100aa4b7fc
CATEGORIES:Conferences - Seminars
DESCRIPTION:Dr. Jörn Jacobsen Senior Research Scientist at the Apple Heal
 th AI team in Zürich. Previously\, he was a postdoc at the Vector Institu
 te and the University of Toronto\, supervised by Rich Zemel and closely co
 llaborating with faculty members David Duvenaud\, Nicolas Papernot and Ro
 ger Grosse. Prior to that\, he was a postdoc in the lab of Matthias Bethge
  in Tübingen and did his Ph.D. at the University of Amsterdam. His work s
 pans a variety of topics: Combining domain knowledge\, physical models an
 d machine learning\, improving robustness of learned models under distribu
 tion shift\, building neural networks with mathematical constraints\, deri
 ving new generative models and analyzing learned representations.\nThere a
 re many methods claiming to help complex machine learning models to genera
 lize when data distributions vary greatly between training and testing con
 ditions. However\, it is also common knowledge that there is no "one-fits-
 all-solution" to this problem. One needs to carefully consider what kind 
 of changes in the data can be expected and design methods accordingly. I w
 ill give an overview of some invariant learning methods I had the pleasure
  to co-develop in this space\, will contextualize them and elaborate on w
 hen I believe they are reasonable to use\, but will also highlight their s
 hortcomings. Finally\, I will present some recent work on using domain kno
 wledge in the form of mechanistic models to overcome some of these shortco
 mings and achieve generalization far beyond the training data.
LOCATION:BM 5202 https://plan.epfl.ch/?room==BM%205202
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR