Write Unit tests

[kobutri]

Figure out a small sample of analytically solvable kmeans datasets.

[idil-tub]

I've did some research and looked through examples.

I've came up with the idea writing unit tests on a smaller size of data (for example 100?) focusing on checking things like:

• if the cluster centers have been computed,
• if labels have been assigned to each point,
• if labels are within the expected range,
• if all labels are distinct,
• if cluster centers are within the range of the data etc.

And some edge cases such as:

• initialization with more clusters than data points,
• single cluster,
• all points are the same,
• completely different clusters,
• no matching clusters

If we also decide to write integration tests to visualize our clusters, we could check the correctness of large datasets with these integration tests.

I've written some mock-up tests that I can show you later in the meeting.

@kobutri @llluww @Kingsleyyy21

[kobutri]

use new api