Repository navigation
v0.3.14
What's new
- Add
metric="nan_euclidean"as an alias for the existing"l2"imputation path under both donor policies. - Support the same aggregation and weight options, numeric missing markers, empty-feature handling, and pandas output.
- Preserve metric parameters for cloning and use L2 for non-Flat index configuration.
This adds a compatible metric name; it does not change the existing L2 algorithm or guarantee identical results to scikit-learn's KNNImputer.