A wavelet-based filtering approach to functional bipartite ranking
Résumé
It is the purpose of this paper to investigate the bipartite ranking task from the perspective of functional data analysis (FDA). Precisely, given a collection of independent copies of a (possibly sampled) random curve X = (X(t))t∊[0,1] taking its values in a function space X, with a locally smooth autocorrelation structure and to which a binary label Y ∊ {−1, +1} is randomly assigned, the goal is to learn a scoring functions: X → R with optimal ROC curve. Based on nonlinear wavelet-based approximation, it is shown how to select compact finite dimensional representations of the input curves in order to build accurate ranking rules, using recent advances in the ranking problem for multivariate data with binary feedback
Mots clés
artificial intelligence
machine learning
Approximation methods
approximation theory
Wavelet transforms
Signal processing
Stochastic processes
Approximation algorithms
Signal processing algorithms
Filtering
wavelet-based filtering approach
functional analysis
filtering theory
functional bipartite ranking
bipartite ranking task
functional data analysis
FDA
ROC curve
nonlinear wavelet-based approximation
input curves
multivariate data
binary feedback
supervised learning
bipartite ranking
ROC optimization
AUC maximization
wavelet analysis