BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Generalization on the unseen domain in real case
DTSTART:20240216T100000
DTEND:20240216T120000
DTSTAMP:20260603T211445Z
UID:ab6f73ce06553cdeb005089fbdc689ec1320e8ec4dcd5b7455edb797
CATEGORIES:Conferences - Seminars
DESCRIPTION:Denys Pushkin\nEDIC candidacy exam\nExam president: Prof. Lén
 aïc Chizat\nThesis advisor: Prof. Emmanuel Abbé\nCo-examiner: Prof. Nico
 las Flammarion\n\nAbstract\nThis write-up summarizes three papers in learn
 ing\ntheory field. The first studies the generalization on the unseen\ndom
 ain in Boolean domain and shows that a number of model\narchitechtures fol
 low the same bias on the unseed. The second\nintroduces a dataset of math 
 problems\, which allows to test\nmodels generalization ability across diff
 erent axes of complexity.\nThe last studies the connection of random featu
 res and nueral\ntangent kernel models with 2-layers neural network.\n\nBac
 kground papers\n1. Emmanuel Abbe\, Samy Bengio\, Aryo Lotfi\, Kevin Rizk 
 "Generalization on the Unseen\, Logic Reasoning and Degree Curriculum" ht
 tps://arxiv.org/pdf/2301.13105.pdf\n2. Behrooz Ghorbani\, Song Mei\, Theo
 dor Misiakiewicz\, Andrea Montanari "Linearized two-layers neural networks
  in high dimension" https://arxiv.org/pdf/1904.12191.pdf\n3. David Saxto
 n\, Edward Grefenstette\, Felix Hill\, Pushmeet Kohli "ANALYSING MATHEMATI
 CAL REASONING ABILITIES OF NEURAL MODELS" https://arxiv.org/pdf/1904.0155
 7.pdf\n 
LOCATION:MA B2 485 https://plan.epfl.ch/?room==MA%20B2%20485
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR