BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:AI Center Seminar - AI Fundamentals series - Dr. Giovanni Marchett
 i
DTSTART:20251111T140000
DTEND:20251111T150000
DTSTAMP:20260501T114832Z
UID:f94297a07cad6cfefb960638831898d36390723223108b841a4a402c
CATEGORIES:Conferences - Seminars
DESCRIPTION:Dr. Giovanni Marchetti\nThe talk is organized by the EPFL AI
  Center as part of the AI fundamentals seminar series.\n\nHosting profess
 or: Prof. Lenka Zdeborova\n\nTitle\nThe Algebraic Geometry of Deep Learnin
 g. \n\nAbstract\nNeural networks parametrize spaces of functions\, someti
 mes referred to as ‘neuromanifolds’. Their geometry is intimately rela
 ted to fundamental machine learning aspects\, such as expressivity\, sampl
 e complexity\, implicit bias\, and training dynamics. For algebraic models
  (e.g.\, networks with a polynomial activation)\, neuromanifolds are (semi
 -) algebraic varieties\, which are the central focus of the field of algeb
 raic geometry. In this talk\, we will provide a general overview of the th
 eory of neuromanifolds of algebraic models\, drawing several connections b
 etween algebraic geometry and deep learning. Along the way\, we will discu
 ss recent results on neuromanifolds of fully-connected networks\, convolut
 ional ones\, and (linear) attention mechanisms. All this lays the foundati
 ons of an emerging discipline that we refer to as Neuroalgebraic Geometry.
 \n\nBio\nI am a postdoctoral researcher at the Department of Mathematics o
 f the Royal Institute of Technology (KTH) in Stockholm\, Sweden.\nI apply 
 tools from pure mathematics (algebra\, geometry\, topology\, ...) to mach
 ine learning and high-dimensional statistics. More specifically\, I am in
 terested in algebro-geometric aspects of deep neural networks\, manifold/r
 epresentation learning\, geometric density estimation\, and topological da
 ta analysis. \n 
LOCATION:ELE 117 https://plan.epfl.ch/?room==ELE%20117 https://epfl.zoom.u
 s/j/68957661036?pwd=najpRJCFurogMPxtNu3Sc2JGQWQX1D.1
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR