rfx-fuse 1.0.0

RFX-Fuse: Breiman and Cutler's Unified ML Engine with GPU Acceleration

RFX-Fuse: Breiman and Cutler's Unified ML Engine

RFX-Fuse delivers classification, regression, unsupervised learning, similarity search, explainability, and outlier detection from a single model.

Installation

pip install rfx-fuse

Prerequisites: CMake 3.12+, Python 3.9+, C++17 compiler, CUDA 12.8+

Quick Example

import RFXFuse as rfx

clf = rfx.RandomForestClassifier(
    ntree=500,
    use_gpu=True,
    compute_importance=True,
    compute_proximity=True
)
clf.fit(X, y)

# OOB error (no separate test set needed)
oob_error = clf.get_oob_error()

# Overall importance
var_imp = clf.feature_importances_()
prox_imp = clf.get_proximity_importance()

# Local importance (per-sample) - e.g., for sample 0
local_var_imp = clf.get_local_importance()[0]      # why was sample 0 predicted this way?
local_prox_imp = clf.get_proximity_importance()[0] # why is sample 0 similar to neighbors?

# Similarity search with explanations
indices, scores, _, feat_idx, feat_imp = clf.get_top_k_similar_with_explanations(0, k=10)

# Outlier detection
outliers, scores = clf.compute_outliers(k=10)

Documentation

• GitHub: https://github.com/chriskuchar/RFX-Fuse
• API Reference: https://github.com/chriskuchar/RFX-Fuse/blob/main/docs/API.md
• Examples: https://github.com/chriskuchar/RFX-Fuse/tree/main/examples

CPU-Only Version

For systems without GPU: pip install rfx-fuse-cpu

License

MIT License