Dimensionality reduction is the process of reducing the number of features in a data set. In a classification problem, the proposed formula allows to sort a set of directions to be used for data projection, according to a score that estimates their capability of discriminating the different data classes. A reduction in the number of features can be obtained by taking a subset of these directions and projecting data on this space. The projecting vectors can be derived from a spectral representation or other choices. If the vectors are eigenvectors of the data covariance matrix, the proposed score is aimed to take the place of the eigenvalues in eigenvector ordering.
Multivariate Direction Scoring for Dimensionality Reduction in Classification Problems / Biagetti, Giorgio; Crippa, Paolo; Falaschetti, Laura; Orcioni, Simone; Turchetti, Claudio. - 56:(2016), pp. 413-423. [10.1007/978-3-319-39630-9_35]
Multivariate Direction Scoring for Dimensionality Reduction in Classification Problems
BIAGETTI, Giorgio;CRIPPA, Paolo
;FALASCHETTI, LAURA;ORCIONI, Simone;TURCHETTI, Claudio
2016-01-01
Abstract
Dimensionality reduction is the process of reducing the number of features in a data set. In a classification problem, the proposed formula allows to sort a set of directions to be used for data projection, according to a score that estimates their capability of discriminating the different data classes. A reduction in the number of features can be obtained by taking a subset of these directions and projecting data on this space. The projecting vectors can be derived from a spectral representation or other choices. If the vectors are eigenvectors of the data covariance matrix, the proposed score is aimed to take the place of the eigenvalues in eigenvector ordering.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
