Rémi Flamary
Site web professionel
English
Français
ZorglangueContact
Filtrage Vaste Marge
Description
Ce code est une implémentation du filtrage vaste marge pour l'étiquetage de séquence 1D ou la classification de pixel en image. Il a été utilisé dans les expérimentations du papier :
R. Flamary, D. Tuia, B. Labbé, G. Camps-Valls, A. Rakotomamonjy, Large Margin Filtering, IEEE Transactions Signal Processing, Vol. 60, N. 2, pp 648-659, 2012.
[Abstract] [BibTeX] [URL] [PDF] [Code]
Abstract: Many signal processing problems are tackled by filtering the signal for subsequent feature classification or regression. Both steps are critical and need to be designed carefully
to deal with the particular statistical characteristics of both
signal and noise. Optimal design of the filter and the classifier are typically aborded in a separated way, thus leading
to suboptimal classification schemes. This paper proposes an
efficient methodology to learn an optimal signal filter and a
support vector machine (SVM) classifier jointly. In particular,
we derive algorithms to solve the optimization problem, prove its
theoretical convergence, and discuss different filter regularizers
for automated scaling and selection of the feature channels. The
latter gives rise to different formulations with the appealing
properties of sparseness and noise-robustness. We illustrate the
performance of the method in several problems. First, linear
and nonlinear toy classification examples, under the presence
of both Gaussian and convolutional noise, show the robustness
of the proposed methods. The approach is then evaluated on
two challenging real life datasets: BCI time series classification
and multispectral image segmentation. In all the examples, large
margin filtering shows competitive classification performances
while offering the advantage of interpretability of the filtered
channels retrieved.
BibTeX:
@article{ieeesp2012,
author = { Flamary, R. and Tuia, D. and Labbé, B. and Camps-Valls, G. and Rakotomamonjy, A.},
title = {Large Margin Filtering},
journal = { IEEE Transactions Signal Processing},
volume = {60},
number = {2},
pages = {648-659},
editor = {},
year = {2012}
} Ce package contient également une version mise à jour de la Toolbox SVM-KM(les classifieurs sont définis par une structure unique au lieu plusieurs matrices), et des fonctions de validation généralistes.
Le package contient:
- SVM-KM : SVM and kernel methods toolbox (voir ici)
- la function svmclass2 qui permet d'apprendre différents SVM
- Des solveurs SVM (libsvm, monqp, svqp2, ...)
- D'autres méthodes de classification: GMM (using netlabs)
Téléchargement
Version courante : 0.9
Téléchargement : FilterSVM.zip
Installation
Version rapide:
- Ajouter tous les dossiers et sous-dossiers au path matlab.
- Executer make.m pour compiler les fichiers mex (libsvm/svqp2)
- Fichier d'entrée : Dataset_Toy/Test_FilterSVM.m
Informations
Hierarchie du code
- 'costs/' : Fonctions de cout SVM
- 'gradients/' : function retournant des gradients SVM (pour filtersvm)
- 'netlab/' : netlab toolbox + fonctions gmmclass/gmmval
- 'optim/' : functions d'optimisation (descente de gradient)
- 'probasolvers/' : fonctions pour des svm probabilistes
- 'svmsolvers/' : plusieurs toolboxes SVM/MKL et le wrapper svmclass2
- 'utils/' : plusieurs fonctions utilitaires
- 'validation/' : boucles de validation/cross-validation
-
'visu/' : functions de visualisation
-
'Dataset_*' : fonctions de test pour diférents jeux de données