BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Title: Why things don’t work — On the foundations of mathemati
 cs and methodological barriers in computations and AI
DTSTART:20210921T141500
DTEND:20210921T150000
DTSTAMP:20260525T033745Z
UID:041ac642a65ddf1f9d122ff1472c3a5e83436b6901e26b99b3b322ae
CATEGORIES:Conferences - Seminars
DESCRIPTION:Anders Hansen leads the Applied Functional and Harmonic Analy
 sis group within the Cambridge Centre for Analysis at DAMTP. He is a R
 eader (Associate Professor) in mathematics at DAMTP\, Professor of Mathem
 atics at the University of Oslo\, a Royal Society University Research Fell
 ow and also a Fellow of Peterhouse.\n\nHis interests include Functional 
 Analysis (applied)\, Foundations of Computations\, Artificial Intelligence
 \, Compressed Sensing\, Optimisation\, Operator/Spectral Theory\, Numerica
 l Analysis\, Computational Harmonic Analysis\, Mathematical Signal Process
 ing\, Sampling Theory\, Inverse Problems\, Medical Imaging\, Geometric Int
 egration\, Operator Algebras\nAbstract: The alchemists wanted to create go
 ld\, Hilbert wanted an algorithm to solve Diophantine equations\, research
 ers want to make deep learning robust in AI\, MATLAB wants (but fails) to 
 detect when it provides wrong solutions to linear programs etc. Why does o
 ne not succeed in so many of these fundamental cases? The reason is typica
 lly methodological barriers. The history of science is full of methodologi
 cal barriers — reasons for why we never succeed in reaching certain goal
 s. In many cases\, this is due to the foundations of mathematics. We will 
 present a new program on methodological barriers and foundations of mathe
 matics\, where — in this talk — we will focus on two basic problems: (
 1) The instability problem in deep learning: Why do researchers fail to pr
 oduce stable neural networks in basic classification and computer vision p
 roblems that can easily be handled by humans — when one can prove that t
 here exist stable and accurate neural networks? Moreover\, AI algorithms c
 an typically not detect when they are wrong\, which becomes a serious issu
 e when striving to create trustworthy AI. The problem is more general\, as
  for example MATLAB's linprog routine is incapable of certifying correct s
 olutions of basic linear programs. Thus\, we’ll answer the following que
 stion: (2) Why are algorithms (in AI and computations in general) incapabl
 e of determining when they are wrong? 
LOCATION:https://epfl.zoom.us/j/63397775674
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR