Resumen |
Cellular genetic algorithms were initially designed
for working on massively parallel machines. However, their
performance can be affected by population size, the kind of
neighborhood and other issues about implementation itself.
Nowadays, GPU cards allow us to execute a task in their
fundamental computing units (called threads), which are
completely independent among them. In addition to use
concurrent threads, there are also memories that help to improve
the performance. In this paper we evaluate the performance of
two models of cellular genetic algorithms, which find the optimal
number of partitions for a data set, using a cluster validation
index as objective function. |