A unifying criterion for unsupervised clustering and feature selection

摘要: Exploratory data analysis methods are essential for getting insight into data. Identifying the most important variables and detecting quasi-homogenous groups of problems interest in this context. Solving such is a difficult task, mainly due to unsupervised nature underlying learning process. Unsupervised feature selection clustering can be successfully approached as optimization by means global heuristics if an appropriate objective function considered. This paper introduces capable efficiently guiding search significant features simultaneously respective optimal partitions. Experiments conducted on complex synthetic suggest that we propose unbiased with respect both number clusters features.