BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//wp-events-plugin.com//6.4.7.2//EN
TZID:Europe/Paris
X-WR-TIMEZONE:Europe/Paris
BEGIN:VEVENT
UID:0-1075@lptms.universite-paris-saclay.fr
DTSTART;TZID=Europe/Paris:20251204T160000
DTEND;TZID=Europe/Paris:20251204T170000
DTSTAMP:20251117T103230Z
URL:https://www.lptms.universite-paris-saclay.fr/seminars/jerome-garnier-b
 run-bocconi/
SUMMARY:MLP@P seminars : Jerome Garnier-Brun (Bocconi) - Salle des séminai
 res du FAST et du LPTMS\, bâtiment Pascal n°530 - 4 Déc 25 16:00
DESCRIPTION:Uncovering Structure: How Neural Networks Learn and Generalize 
 from Tree-based Data\nJerome Garnier-Brun\nBocconi University\, Milan\n\nS
 eminar of the Series MLP@P (Machine Learning Physics @ Plateau)\, joint wi
 th LISN and IPhT.\nWhere: LPTMS\, Salle des Séminaires (1° étage)\n\nSt
 atistical-physics approaches have provided key insights into the functioni
 ng of neural networks\, yet most analyses assume high-dimensional random d
 ata. In contrast\, real-world data possess rich underlying structure that 
 likely shapes how learning and generalization unfold. In an attempt to bri
 dge this gap\, this talk will focus on models trained on tree-based data\,
  a setting where Bayes-optimal performance can importantly still be comput
 ed exactly. By introducing a controlled filtering procedure that tunes the
  degree of correlation in the data\, we first probe how transformers progr
 essively uncover structure in both supervised and self-supervised inferenc
 e tasks. The results reveal a hierarchical discovery of correlations—fir
 st in time\, during training\, and in space\, across attention layers—wh
 ich closely mirrors the exact inference algorithm. In a second part\, we t
 urn to generative diffusion models\, where the same controlled data model 
 exposes a novel biased generalization regime that precedes overt overfitti
 ng. There\, access to Bayes-optimal benchmarks allows a precise characteri
 zation of when and how this bias emerges\, suggesting new directions for o
 ptimizing diffusion schedules.
CATEGORIES:MLP@P
LOCATION:Salle des séminaires du FAST et du LPTMS\, bâtiment Pascal n°53
 0\, rue André Riviere\, Orsay\, 91405\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=rue André Riviere\, Orsay\
 , 91405\, France;X-APPLE-RADIUS=100;X-TITLE=Salle des séminaires du FAST 
 et du LPTMS\, bâtiment Pascal n°530:geo:0,0
END:VEVENT
BEGIN:VTIMEZONE
TZID:Europe/Paris
X-LIC-LOCATION:Europe/Paris
BEGIN:STANDARD
DTSTART:20251026T020000
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
END:STANDARD
END:VTIMEZONE
END:VCALENDAR