Feature Selection Using Counting Grids: Application to Microarray Data

Abstract

Abstract. In this paper a novel feature selection scheme is proposed, which exploits the potentialities of a recent probabilistic generative model, the Counting Grid. This model is able to cluster together similar observations, highlighting the compactness of a class and its underlying structure. The proposed feature selection scheme is applied to the expression microarray scenario, a peculiar context with very few patterns and a huge number of features. Experiments on benchmark datasets show that the proposed approach is effective and stable, assessing state-of-the-art classification accuracies.

Feature Selection for High-Dimensional Data

“…To overcome this problem, Lovato et al [69] presented a novel feature selection scheme, based on the Counting Grid (GC) model, which can measure and consider the relation and the influence between genes. Most of the gene selection techniques are based on the assumption of the independence between genes (actually a typical approach is to rank them individually).…”

Feature Selection in DNA Microarray Classification

“…However, it is well known that genes interact with each other through gene regulative networks. To overcome this problem, Lovato et al [69] presented a novel feature selection scheme, based on the Counting Grid (GC) model, which can measure and consider the relation and the influence between genes.…”

“…In fact, reviewing the recent literature one can find examples of the four methods described above. k-fold cross-validation is a common choice [42,44,48,57,58,61,68], as is holdout validation [43,46,52,55,59,63,69]. Bootstrap sampling was used less [50,60,66], probably due to its high computational cost, and there are also some representatives of leave-one-out cross-validation [62,64].…”

Exploiting Geometry in Counting Grids

“…The starting point is a microarray gene expression matrix, where the element at position (i, j) represents the expression level of the i − th gene in the j − th subject/sample. Methods based on counting values (as CG and topic models) have been recently and successfully applied in this context (see, e.g., [16,17,11]). This is possible if we establish an analogy between a word-document pair and a gene-sample pair; it seems reasonable to interpret samples as documents and genes as words.…”

“…In particular, with CG an ordering procedure between BoWs is introduced by allowing BoWs to lie in an n-dimensional grid structure. Such approach has already shown its benefits on document retrieval, 2D scene classification, and microarray expression classification [1,11]. In all these applications, the classification stage has been computed by standard maximum likelihood scheme, or by employing discriminative classifiers like Support Vector Machine (SVM) with generative kernels, nevertheless without taking into account the peculiar geometry of the model.…”