N. Ancona, A. D' Addabbo, S. Liuni, G. Pesole, and R. Maglietta (Italy)
Support Vector Machine, Classification, Computational ge nomics
In this paper we compare the performances of two super vised learning techniques for cancer classification by DNA microarray data: Regularized Least Squares (RLS) classi fiers, originally proposed in regularization theory, and Sup port Vector Machines (SVM). For a detailed comparison of SVM and RLS performances, we analyze data sets with dif ferent numbers of genes selected by statistical methods. We show that RLS classifiers have comparable performances to the one of SVM classifiers as the Leave-One-Out (LOO) error evaluated on two different data sets shows. The main advantage of RLS machines is that for solving a classifica tion problem they solve a linear system of order equal to the number of training examples. Moreover, RLS machines al low to get an exact measure of the LOO error with just one training.
Important Links:
Go Back
