kde-classifier 0.1.0

Kernel Density Estimation classifier with scikit-learn style API, based on Yang Liu et al. (2020).

KDEClassifier

A pip-installable distribution of KDEClassifier introduced by:
Yang Liu, Hongxia He, Zhi-Xiong Xiao, Anquan Ji, Jian Ye, Qifan Sun, Yang Cao.
See: https://pubmed.ncbi.nlm.nih.gov/33313714/

Code adapted from: http://cao.labshare.cn/KDEClassifier/index.html

This repository provides two implementations:

• kde_classifier.KDEClassifier — An enhanced version of the original implementation; vectorized and sklearn-compatible.
• tests/KDEClassifier_orig.py — the original reference implementation (kept for comparison and tests).

---

Installation

pip install kde-classifier

Usage example

from kde_classifier import KDEClassifier

clf = KDEClassifier(mybandwidth=0.7)
clf.fit(X_train, y_train)
probs = clf.predict_proba(X_test)
preds = clf.predict(X_test)

predict_proba always returns a 2-D array shaped (n_samples, n_classes) (even when n_samples==0).

---

Tests

A pytest test file is provided under tests/ that compares the original implementation with the new one on synthetic datasets. Run tests with:

pytest -q

---