Identifiable up to rearrangement

[Hamster_on_wheels]

Something like transform.ordered might work. But I need to sort both row and columns by the ordering of row sums.

Divergences and label-switching in a Dirichlet process mixture model

TL;DR: I’m attempting to perform non-parametric clustering of circular data using a Dirichlet process mixture of von Mises distributions. I’m trying to fix label-switching using transforms.ordered, but my chains contain only divergent samples. Details: The observed data are stimulus values (colour) from a cognitive psychology experiment. Each trial of the experiment used 6 stimuli sampled from a uniform distribution. For our analysis we would like to characterize possible clusters of stimuli t…