Different variants of Normalized EM Algorithm for Gene Expression Data
The actual clustering methods of directional data are commonly performed in considering data belonging to the unit radius sphere. In certain situations, assuming that the radius equal to one is not appropriated. In this work we propose to estimate it to be greater than one. We then suggest to apply a normalized EM algorithm and derive different variants. This procedure has shown an important influence on the clustering results of gene expression data.
22nd International Workshop on Database and Expert Systems Applications The 2nd International Workshop on Biological Knowledge Discovery and Data Mining - DEXA '11conference proceeding 2011-09-01