Centroid movement and transferences: alternatives for generating neighbors in simulated annealing

Main Article Content

Jeffry Chavarría-Molina
Juan José Fallas-Monge

Abstract

In this paper, two different strategies to generate neighbors in the simulated annealing algorithm were compared. The first idea is based on transferring objects between classes. The other strategy moves artificial objects (called centroids) which represent the clusters. A comparison was developed using the combinatorial optimization problem of quantitative data clustering. This problem is presented as the minimization of W(P), which represents the within-inertia as a function of the partition P. Finally, this allowed comparison of the algorithms’ performance in several data sets. 

Article Details

How to Cite
Chavarría-Molina, J., & Fallas-Monge, J. J. (2016). Centroid movement and transferences: alternatives for generating neighbors in simulated annealing. Tecnología En Marcha Journal, 29(5), pág. 64–76. https://doi.org/10.18845/tm.v29i5.2586
Section
Artículo científico