Table 4 Glossary

${\alpha }_E^{\ast }$

The value of the DDP leading to the best separation of classes in terms of the Euclidean distance

${\alpha }_M^{\ast }$

The value of the DDP leading to the best separation of classes in terms of the Manhattan distance in original feature space

${\alpha }_U^{\ast }$

The value of the DDP giving the largest antiredundancy in PC space

${\alpha }_{MP}^{\ast }$

The value of the DDP leading to the best separation of classes in terms of the Manhattan distance in PC space

α

A variable representing the DDP

Antiredundancy

A parameter opposite to redundancy in terms of quality and thus is to be maximized along with relevance

DDP

Degree of differential prioritization, which controls the balance between the two requirements in feature selection (maximizing relevance and maximizing antiredundancy)

Equal-priorities scoring methods

Filter-based feature selection techniques in which the predictor set scoring method places equal importance on relevance and redundancy as criteria in forming the predictor set

G

Number of genes in each group of marker genes, a parameter set during the generation of toy datasets

K

Number of classes in the dataset

m

Class size (number of samples per class), a parameter set during the generation of toy datasets

N

Number of genes in the dataset

OVA

One-vs.-all

P

Predictor set size, i.e., number of genes selected into the predictor set

PCA

Principal component analysis

PW

Pairwise

Rank-based selection or rank-based techniques

Filter-based feature selection techniques in which relevance is the sole criterion in forming the predictor set

Redundancy

The redundancy in a predictor set indicates the amount of similarity among the members of the predictor set

Relevance

The ability to distinguish among different classes

S _α

The predictor set found using a DDP value of α