Evolutionary Monte Carlo (EMC) Methods for Clustering
Description
Evolutionary Monte Carlo methods for clustering, temperature
ladder construction and placement. This package implements methods
introduced in Goswami, Liu and Wong (2007) .
The paper above introduced probabilistic genetic-algorithm-style crossover
moves for clustering. The paper applied the algorithm to several clustering
problems including Bernoulli clustering, biological sequence motif
clustering, BIC based variable selection, mixture of Normals clustering,
and showed that the proposed algorithm performed better both as a sampler
and as a stochastic optimizer than the existing tools, namely, Gibbs sampling,
``split-merge'' Metropolis-Hastings algorithm, K-means clustering, and the
MCLUST algorithm (in the package 'mclust').