BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//wp-events-plugin.com//7.2.3.1//EN
TZID:Europe/Paris
X-WR-TIMEZONE:Europe/Paris
BEGIN:VEVENT
UID:2298@i2m.univ-amu.fr
DTSTART;TZID=Europe/Paris:20180413T100000
DTEND;TZID=Europe/Paris:20180413T110000
DTSTAMP:20180329T080000Z
URL:https://www.i2m.univ-amu.fr/evenements/spectral-learning-of-weighted-a
 utomata-multitask-setting-and-nonlinear-extension/
SUMMARY: (...): Spectral learning of weighted automata: multitask setting a
 nd nonlinear extension
DESCRIPTION:: Structured objects such as strings\, trees and graphs are ubi
 quitous in data science but learning functions defined over such objects c
 an be a tedious task. Weighted automata (WA) are powerfultools that can ef
 ficiently model such functions and are thus particularly relevant for mach
 ine learning. In particular\, the spectral learning algorithm offers an ef
 ficient way to learn WA which comes with strong theoretical guarantees. In
  this talk\, I will present two recent extensions of the spectral learning
  algorithm. The first one addresses the multitask learning problem for WA:
  how can one leverage relatedness between two or more WAs to learn more ef
 ficiently? After introducing the novel model of vector-valued WA (which co
 nveniently helps formalizing this problem)\, I will show how the spectral 
 learning algorithm can be extended to vector-valued WA and showcase the be
 nefits of this approach with experiments on a natural language modeling ta
 sk. In the second part of the talk I will discuss connections between WA a
 nd recurrent neural networks and present a non-linear extension of WA alon
 g with a learning algorithm. This learning algorithm can be seen as a non-
 linear counterpart to the classical spectral learning algorithm\, where th
 e factorization of the Hankel matrix is replaced by an auto-encoder networ
 k and each transition function is realized by a feed-forward network.http:
 //cs.mcgill.ca/~grabus/
END:VEVENT
BEGIN:VTIMEZONE
TZID:Europe/Paris
X-LIC-LOCATION:Europe/Paris
BEGIN:DAYLIGHT
DTSTART:20180325T030000
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
END:DAYLIGHT
END:VTIMEZONE
END:VCALENDAR