W-k-NN 0.0.3

A W-k-NN implementation with PyKeOps

W-K-NN: Weighted-k-Nearest-Neighbours

A GPU implementation of the optimal [1] W-k-NN classifier based on KeOps [2]

Installation

Using pip: pip install W-k-NN

Usage

classifier file is WKNN/wknn_clf.py

from WKNN.wknn_clf import WKNNClassifier

model = WKNNClassifier().fit(X,y)
y_pred = model.predict(X_test)

Tutorial

See compare-classifiers-synth-2D.ipynb for comparing the implementation with state-of-the-art methods.
See wknn-tuto-2d.ipynb notebook for a tutorial (To Do).

References

[1] Richard J. Samworth, Optimal weighted nearest neighbour classifiers, 2012, doi:10.1214/12-AOS1049
[2] Charlier, B., Feydy, J., Glaunès, J. A., Collin, F.-D. & Durif, G. Kernel Operations on the GPU, with Autodiff, without Memory Overflows. Journal of Machine Learning Research 22, 1–6 (2021).

Author & License

Copyright UCA/CNRS/Inria
Contributor(s): Cedric Dubois 2022

This software is governed by the CeCILL license under French law and abiding by the rules of distribution of free software. You can use, modify and/ or redistribute the software under the terms of the CeCILL license as circulated by CEA, CNRS and INRIA at the following URL "http://www.cecill.info".