On Sat, May 11, 2013 at 9:43 AM, Matthew McClain <mattmccla...@gmail.com>wrote:
> This constraint can be > removed by characterizing each cluster by the mean and covariance of its > samples, and using maximum likelihood in place of the distance measurement > for assigning clusters to samples. > Just a note that ordinary k-means doesn't work well with variable covariance. You need some form of regularization. The Dirichlet clustering in Mahout provides on such method for doing this.