Added smooth and non-smooth prediction functions with tests and comments

[ClaudioZeni]

Proposed changes

Added two predict functions to Clustering class.
One follows the Density Peaks criteria and assigns cluster labels based on the label of the closest neighbour of higher density.
Such density can be computed using PAk or kstarNN. This function predicts only labels and assigns probabilities that are always 1 or 0.
The second function employs a smoothed majority rule to assign labels and label probabilities.
Points in the kstar neighbourhood are weighted according to a normalized inverse square distance from the out-of-set point.
Label is then predicted according to weighted majority rule, and probabilities are given by the weighted population of each label.

Types of changes

[New Feature] Allows using Clustering to predict the labels and probability of each label for out-of-set points.

Checklist

• [ x ] Lint and unit tests pass locally with my changes
• [ x ] I have added tests that prove my fix is effective or that my feature works

[imacocco]

I think this PR can be accepted and merged. Is there some more work to do maybe?

[AldoGl]

One small problem is that "density_est" does not appear in the docstring.

On the algorithm I think also @alexdepremia could give his opinion

[ClaudioZeni]

On my side the PR can be merged, I guess it's up to the reviewers

…

On Wed, Jul 6, 2022 at 10:38 AM Iuri Macocco ***@***.***> wrote: I think this PR can be accepted and merged. Is there some more work to do maybe? — Reply to this email directly, view it on GitHub <#61 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AHJWOAJB4R4FNNJ4TMVDAN3VSVHZPANCNFSM5WROLGXQ> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[alexdepremia]

Hi, sorry for the late reply. I think it can be merged, but we should clearly state that it is an experimental feature (it has not been published).