BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:IEM Seminar Series: Machine Learning for plant identification and 
 other challenges from citizen science
DTSTART:20230403T100000
DTEND:20230403T110000
DTSTAMP:20260405T232402Z
UID:0996f4de38652f6e833d3ac3475dd1377a8851566f574f9f6cd7ecbc
CATEGORIES:Conferences - Seminars
DESCRIPTION:Prof Joseph Salmon\nUniversité de Montpellier\, France\nAbstr
 act\nIn supervised learning --- for example for image classification --- m
 odern massive datasets are usually labeled by a crowd of "workers".\nAn ex
 ample of interest is the case of citizen science\, where data is potential
 ly provided and annotated by volunteers. In particular\, we will describe 
 the learning challenges related to an application such as Pl@ntnet\, aimin
 g at plant species recognition with a simple cell phone leveraging both ex
 pert and amateur botanists.\nThe labels obtained in such a framework must 
 be aggregated to improve learning. \nThis aggregation step is usually bas
 ed on a confidence score per annotator.\nHowever\, these annotator-centric
  approaches do not take into account the ambiguity of each task.\nSome inh
 erently ambiguous tasks may even mislead expert workers\, which could pote
 ntially hinder the learning step.\nWe will present an adaptation - to part
 icipatory learning scenarios - of the AUM (Area Under the Margin)\, a tool
  for identifying ambiguous tasks\, which allows for improved generalizatio
 n or calibration performance.\n\nBio\nJoseph Salmon is a full professor at
  Univ. Montpellier since 2018.\nPrior to joining Univ. Montpellier\, he wa
 s a visiting professor at UW from February to August 2018\, and an assista
 nt professor at Télécom ParisTech from 2012 to 2018.\nHis research inter
 ests include cooperative learning\, convex optimization\, sparse regressio
 n models and inverse problems from imaging science.\nHis recent contributi
 ons have aimed at improving machine learning techniques for plant identifi
 cation\, leveraging crowdsourcing and citizen science.\nPrior works have f
 ocused on image denoising (in particular for low photon counts)\, on speed
 ing-up standard Lasso solvers and on handling the noise structure to impro
 ve signal estimation.\nAs an associate member of the INRIA Parietal Team\,
  he was contributing to applying his work to brain imaging challenges.\n 
LOCATION:MED 0 1418 https://plan.epfl.ch/?room==MED%200%201418 https://epf
 l.zoom.us/j/65042188582
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR