BEGIN:VCALENDAR VERSION:2.0 PRODID:-//wp-events-plugin.com//7.4.0.1//EN TZID:Europe/Paris X-WR-TIMEZONE:Europe/Paris BEGIN:VEVENT UID:138@isdm.umontpellier.fr DTSTART;TZID=Europe/Paris:20250310T153000 DTEND;TZID=Europe/Paris:20250310T153000 DTSTAMP:20260602T130233Z URL:https://isdm.umontpellier.fr/events/distributional-matrix-completion-v ia-nearest-neighbors-in-the-wasserstein-space-2/ SUMMARY:Distributional Matrix Completion via Nearest Neighbors in the Wasse rstein Space DESCRIPTION:Campus St Priest (860 Rue Saint Priest 34095 Montpellier Cedex 5)\, bat. 5\, Room: 02.124\nMachine Learning in Montpellier\, Theory &\ ; Practice\nJacob Feitelberg\n\nWe study the problem of distributional mat rix completion: Given a sparsely observed matrix of empirical distribution s\, we seek to impute the true distributions associated with both observed and unobserved matrix entries. This is a generalization of traditional ma trix completion where the observations per matrix entry are scalar-valued. To do so\, we utilize tools from optimal transport to generalize the near est neighbors method to the distributional setting. Under a suitable laten t factor model on probability distributions\, we establish that our method recovers the distributions in the Wasserstein metric. We demonstrate thro ugh simulations that our method (i) provides better distributional estimat es for an entry compared to using observed samples for that entry alone\, (ii) yields accurate estimates of distributional quantities such as standa rd deviation and value-at-risk\, and (iii) inherently supports heterosceda stic distributions. In addition\, we demonstrate our method on a real-worl d quarterly earnings predictions dataset. We also prove novel asymptotic r esults for Wasserstein barycenters over one-dimensional distributions.\n\n LabéliséHallesIA ATTACH;FMTTYPE=image/jpeg:https://isdm.umontpellier.fr/wp-content/uploads/ 2025/02/ml-mpt-1.jpg END:VEVENT BEGIN:VTIMEZONE TZID:Europe/Paris X-LIC-LOCATION:Europe/Paris BEGIN:STANDARD DTSTART:20241027T020000 TZOFFSETFROM:+0200 TZOFFSETTO:+0100 TZNAME:CET END:STANDARD END:VTIMEZONE END:VCALENDAR