search

mariaderrico / DPA

The DPA package is the scikit-learn compatible implementation of the Density Peaks Advanced clustering algorithm. The algorithm provides robust and visual information about the clusters, their statistical reliability and their hierarchical organization.
BSD 3-Clause "New" or "Revised" License
27 stars 9 forks source link

Accept callable metric #1

Closed giovannidoni closed 4 years ago

giovannidoni commented 4 years ago

A callable should be a viable distance metric to be used as per the docstring, however the constructor raises an error if a callable is passed. Added a few costmetic changes.

mariaderrico commented 4 years ago

Thanks for adding a check on the metric for the case it is callable, and the corresponding test function.

Among the cosmetic changes I cannot accept those removing self.k_max_ and self.dim_: the attributes self.k_max and self.dim should not be modified, and thus the new public attributes self.k_max_ and self.dim_ are defined in the fit function. Indeed the test_common.py test fails with the following errors:

E           AssertionError: Estimator DensityPeakAdvanced should not change or mutate  the parameter dim from None to 2 during fit.

E       AssertionError: Estimator changes public attribute(s) during the fit method. Estimators are only allowed to change attributes started or ended with _, but dim changed

=============================================== short test summary info ================================================
FAILED test_common.py::test_scikitlearn_compatible_estimator[DensityPeakAdvanced()-check_estimators_overwrite_params]
FAILED test_common.py::test_scikitlearn_compatible_estimator[DensityPeakAdvanced()-check_dont_overwrite_parameters]
============================================= 2 failed, 87 passed in 1.07s =============================================

It would also be better to have separate commits for the callable metric and the cosmetic changes.

giovannidoni commented 4 years ago

See pr #2