BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:Improving PLS with lasso shrinkage\, using the dual-SPLS package
DTSTART:20220627T093000Z
DTEND:20220627T100000Z
DTSTAMP:20240303T024800Z
UID:indico-contribution-305@conferences.enbis.org
DESCRIPTION:Speakers: Louna Alsouki (IFPEN)\, Laurent Duval (IFPEN)\, Fra
nçois Wahl (IFPEN)\, Clément Marteau (Institut Camille Jordan)\, Rami El
Haddad (Université Saint Joseph de Beyrouth)\n\nIn analytical chemistry\
, high dimensionality problems in regression are generally solved using tw
o dimension reduction techniques: projection methods\, one of which is PLS
or variable selection algorithms\, as in lasso. Sparse PLS combines both
approaches by adding a variable selection step to the PLS dimension reduct
ion scheme. However\, in most existing algorithms\, interpretation of the
remaining coefficients is usually doubtful.\n\nWe conceived a generalizati
on of the classical PLS1 algorithm\, i.e. when the response is one-dimensi
onal\, the dual-SPLS\, aiming at providing sparse coefficients for good in
terpretation while maintaining accurate predictions. Dual-SPLS is based on
the reformulation of the PLS1 problem as a dual L2 norm procedure. Varyin
g the underlying norm introduces regularization aspects in PLS1 algorithm.
\n\nChoosing a mix of L1 and L2 norms brings shrinkage in the selection of
each PLS component in an analogous way to the lasso procedure\, depending
on a parameter $\\nu$. The method elaborated in dual-SPLS adaptively sets
the value of $\\nu$ according to the amount of desired shrinkage in a use
r-friendly manner. \n\nIndustrial applications of this algorithm provide a
ccurate predictions while extracting pertinent localization of the importa
nt variables.\n\nMoreover\, extending the underlying norm to cases with he
terogeneous data is straightforward.\n\nWe present here some applications
(simulated and real industrial cases) of these procedures while using a de
dicated toolbox in R: dual.spls package.\n\nhttps://conferences.enbis.org/
event/18/contributions/305/
LOCATION:EL6
RELATED-TO:indico-event-18@conferences.enbis.org
URL:https://conferences.enbis.org/event/18/contributions/305/
END:VEVENT
END:VCALENDAR