BEGIN:VCALENDAR VERSION:2.0 PRODID:-//132.216.98.100//NONSGML kigkonsult.se iCalcreator 2.20.4// BEGIN:VEVENT UID:20250810T035254EDT-8319ZDExPj@132.216.98.100 DTSTAMP:20250810T075254Z DESCRIPTION:Title: The Mathematical Foundations of Deep Learning: From Rati ng Impossibility to Practical Existence Theorems.\n\nAbstract: Deep learni ng is having a profound impact on industry and scientific research. Yet\, while this paradigm continues to show impressive performance in a wide var iety of applications\, its mathematical foundations are far from being wel l established. In this talk\, I will present recent developments in this a rea by illustrating two case studies.\n \n First\, motivated by applications in cognitive science\, I will present “rating impossibility' theorems. Th ey identify frameworks where deep learning is provably unable to generaliz e outside the training set for the seemingly simple task of learning ident ity effects\, i.e. classifying whether pairs of objects are identical or n ot.Second\, motivated by applications in scientific computing\, I will ill ustrate “practical existence' theorems. They combine universal approximati on results for deep neural networks with compressed sensing and high-dimen sional polynomial approximation theory. As a result\, they yield sufficien t conditions on the network architecture\, the training strategy\, and the number of samples able to guarantee accurate approximation of smooth func tions of many variables.Time permitting\, I will also discuss work in prog ress and open questions.To receive the announcements with the Zoom link\, please subscribe once to the lists of your choice\n\nWeb site : http://crm .umontreal.ca/colloque-sciences-mathematiques-quebec/index.html#csmq\n\nHy bride - CRM\, Salle / Room 5340\, Pavillon André Aisenstadt / Zoom (voir s urvol/ see overview)\n DTSTART:20230210T203000Z DTEND:20230210T213000Z SUMMARY:Simone Brugiapaglia (Concordia University) URL:/mathstat/channels/event/simone-brugiapaglia-conco rdia-university-345751 END:VEVENT END:VCALENDAR