Cluster assignments sometimes produces random numbers

[charliematteo]

I realised that despite N_clusters is correctly identified, a few points are assigned as totally random integers, i.e. not between 0 (or -1) and N_clusters-1.

This happens if at the previous step I compute density PAk, not with kstarNN.

I suspect this is due to something in the cython_clustering.pyx. It looks like some points are left out of the assignment, so, since the cluster assignment array is initialised as an empty array
Last_cls = np.empty(Nele, dtype=int)
I think it just takes random values for those points. This is however just my hypothesis.

I attach a screenshot from my jupyter notebook.
The previous command is
d.compute_density_PAk()

d is an instance of the Clustering class: d = Clustering(X, maxk=200, verbose=True, period=periodicity)