Features/599 kmeans and friends

[Cdebus]

Description

Implementation of kmedians and kmedoids-version (kmedians snaped to nearest data point). Along that line, kmeans was refactored by introducing a base class _kcluster, that handles centroid initialization, predict function and assigning data points to the respective cluster centers. Kmeans, kmedians and kmedoids are derivatives of _kcluster with specific metrics (euclidean, manhattan) and different implementations of the centroid updating function.

Issue/s resolved: #409 #599

Changes proposed:

• Introduce _kcluster.py and move functions initialize_centroids, fit_to_cluster (=assign_to_cluster) and predict there
• Implement manhattan distance in spatial.distances
• Implement kmedians as _kcluster with manhattan distance and median calculation for updating centroids
• Implement kmedoids as variation of kmedians, that assigns the updated cluster centers to the data point lying closest to the actual median in each iteration
• Implement an example on spherical clusters to see precision and differences between the three k-clustering algorithms
• some minor bugfixes

Type of change

• New feature (non-breaking change which adds functionality)

Due Diligence

• All split configurations tested : No Unittest for kmedians and kmedoids for unsplit data, as this causes transient errors in the unit tests. However, functionality was checked in applications.
• Multiple dtypes tested in relevant functions
• Documentation updated (if needed)
• Updated changelog.md under the title "Pending Additions"

Does this change modify the behaviour of other functions? If so, which?

no

[codecov]

Codecov Report

Merging #634 into master will decrease coverage by 0.12%.
The diff coverage is 92.70%.

@@            Coverage Diff             @@
##           master     #634      +/-   ##
==========================================
- Coverage   97.40%   97.27%   -0.13%     
==========================================
  Files          79       84       +5     
  Lines       16126    16535     +409     
==========================================
+ Hits        15708    16085     +377     
- Misses        418      450      +32     

Impacted Files	Coverage Δ
heat/core/_operations.py	92.73% <ø> (ø)
heat/cluster/kmedians.py	70.58% <70.58%> (ø)
heat/cluster/kmedoids.py	76.19% <76.19%> (ø)
heat/cluster/_kcluster.py	88.78% <88.78%> (ø)
heat/cluster/__init__.py	100.00% <100.00%> (ø)
heat/cluster/kmeans.py	96.77% <100.00%> (+5.86%)	⬆️
heat/cluster/tests/test_kmeans.py	100.00% <100.00%> (ø)
heat/cluster/tests/test_kmedians.py	100.00% <100.00%> (ø)
heat/cluster/tests/test_kmedoids.py	100.00% <100.00%> (ø)
heat/core/__init__.py	100.00% <100.00%> (ø)
... and 15 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 6629499...ef92850. Read the comment docs.