EscalanteB., AlbertoN. and Wiskott, Prof. Dr. Laurenz (2013) How to Solve Classification and Regression Problems on HighDimensional Data with a Supervised Extension of Slow Feature Analysis. [Preprint]
Full text available as:

PDF
1563Kb 
Abstract
Supervised learning from highdimensional data, e.g., multimedia data, is a challenging task. We propose an extension of slow feature analysis (SFA) for supervised dimensionality reduction called graphbased SFA (GSFA). The algorithm extracts a labelpredictive lowdimensional set of features that can be postprocessed by typical supervised algorithms to generate the ﬁnal label or class estimation. GSFA is trained with a socalled training graph, in which the vertices are the samples and the edges represent similarities of the corresponding labels. A new weighted SFA optimization problem is introduced, generalizing the notion of slowness from sequences of samples to such training graphs. We show that GSFA computes an optimal solution to this problem in the considered function space, and propose several types of training graphs. For classiﬁcation, the most straightforward graph yields features equivalent to those of (nonlinear) Fisher discriminant analysis. Emphasis is on regression, where four different graphs were evaluated experimentally with a subproblem of face detection on photographs. The method proposed is promising particularly when linear models are insufficient, as well as when feature selection is difficult.
Item Type:  Preprint 

Keywords:  Slow feature analysis, feature extraction, classiﬁcation, regression, pattern recognition, training graphs, nonlinear dimensionality reduction, supervised learning, highdimensional data, implicitly supervised, image analysis. 
Subjects:  Computer Science > Machine Learning Computer Science > Machine Vision Computer Science > Neural Nets 
ID Code:  8966 
Deposited By:  EscalanteB., AlbertoN. 
Deposited On:  04 May 2013 23:24 
Last Modified:  04 May 2013 23:24 
Metadata
 ASCII Citation
 Atom
 BibTeX
 Dublin Core
 EP3 XML
 EPrints Application Profile (experimental)
 EndNote
 HTML Citation
 ID Plus Text Citation
 JSON
 METS
 MODS
 MPEG21 DIDL
 OpenURL ContextObject
 OpenURL ContextObject in Span
 RDF+NTriples
 RDF+N3
 RDF+XML
 Refer
 Reference Manager
 Search Data Dump
 Simple Metadata
 YAML
Repository Staff Only: item control page