KMeans check_array performance
Open
- Dominant language
- Python
- Stars
- 951
- Forks
- 262
- PR merge metrics
- No merged PRs in 30d
Description
In KMeans, we currently do a full scan of the data for nan / inf values before starting the clustering.
I wonder if that can be
a.) optionally disabled
b.) pushed down to https://github.com/dask/dask-ml/blob/56714b4a8749d0b0d1948de443bcce43dc60354e/dask_ml/cluster/k_means.py#L573, and run on just the first iteration.
Contributor guide
Assessment
This issue has not been assessed yet.