n_init

The number of times k-means (or a Gaussian mixture) runs from different starting points, keeping the best run. A value such as `n_init=10` for k-means costs ten times the work of a single run but makes a poor start less likely.