Including cluster mean as a cluster feature
Opened this issue · 0 comments
MichaelKUL commented
Hi David, I have been implementing TICC on unlabeled data to identify interesting clusters for my thesis. Am I correct in stating that cluster means are not a discriminating feature for a cluster? In your paper you state that you generated data for every cluster with mean zero. Although I can standardize my real-life dataset, I can not be certain my clusters are centered around 0.
What do you suggest if cluster means are important? Is it possible to define clusters by their mean and covariance (µ, Θ)?