update n_init for kmeans

[md-shafiul-alam]

• For kmeans++, the default n_init=1 https://github.com/scikit-learn/scikit-learn/blob/main/sklearn/cluster/_kmeans.py#L346. Probably makes sense to benchmark the default behavior.

[Alexsandruss]

@md-shafiul-alam

Change normalize to false

But why? KMeans inertia is sensitive to normalization

[md-shafiul-alam]

@md-shafiul-alam

Change normalize to false

But why? KMeans inertia is sensitive to normalization

Only reason is that stock scikit learn does not have any option for normalization.

[Alexsandruss]

data:preprocessing_kwargs:normalize is not argument of an estimator

[md-shafiul-alam]

data:preprocessing_kwargs:normalize is not argument of an estimator

I understand, but is there a specific reason we normalize by default? To me seems like a redundant step.

[Alexsandruss]

data:preprocessing_kwargs:normalize is not argument of an estimator

I understand, but is there a specific reason we normalize by default? To me seems like a redundant step.

Normalization is required for algorithms sensitive to it for better quality metrics and representation of user cases.

[md-shafiul-alam]

@Alexsandruss Do you suggest any more changes?