Leave-one-out evaluation of the nearest feature line and the rectified nearest feature line segment classifiers using Multi-core architectures

Abstract

In this paper we present the parallelization of the leave-one-out test: areproducible test that is, in general, computationally expensive. Paral-lelization was implemented on multi-core multi-threaded architectures, us-ing the Flynn Single Instruction Multiple Data taxonomy. This techniquewas used for the preprocessing and processing stages of two classificationalgorithms that are oriented to enrich the representation in small samplecases: the nearest feature line (NFL) algorithm and the rectified nearestfeature line segment (RNFLS) algorithm. Results show an accelerationof up to 18.17 times with the smallest dataset and 29.91 times with thelargest one, using the most costly algorithm (RNFLS) whose complexityisO(n4). The paper also shows the pseudo-codes of the serial and parallel algorithms using, in the latter case, a notation that describes the way theparallelization was carried out as a function of the threads.