Feature selection has been studied widely in the literature. However, theefficacy of the selection criteria for low sample size applications isneglected in most cases. Most of the existing feature selection criteria arebased on the sample similarity. However, the distance measures becomeinsignificant for high dimensional low sample size (HDLSS) data. Moreover, thevariance of a feature with a few samples is pointless unless it represents thedata distribution efficiently. Instead of looking at the samples in groups, weevaluate their efficiency based on pairwise fashion. In our investigation, wenoticed that considering a pair of samples at a time and selecting the featuresthat bring them closer or put them far away is a better choice for featureselection. Experimental results on benchmark data sets demonstrate theeffectiveness of the proposed method with low sample size, which outperformsmany other state-of-the-art feature selection methods.
展开▼
