Fast-MINE

A Numba-accelerated Python implementation of MINE (Maximal Information-based Nonparametric Exploration) statistics, including MIC (Maximal Information Coefficient).

This package provides a drop-in replacement for the C-based minepy implementation, optimized for performance using JIT compilation via Numba.

The logic and algorithms in this package are derived from the original minepy implementation and the methodologies described in:

• David N. Reshef, Yakir A. Reshef, Hilary K. Finucane, et al. "Detecting Novel Associations in Large Data Sets." Science 334, 1518 (2011). DOI: 10.1126/science.1205438
• Davide Albanese, Michele Filosi, Roberto Visintainer, Samantha Riccadonna, Giuseppe Jurman, and Cesare Furlanello. "minerva and minepy: a C engine for the MINE suite and its R, Python and MATLAB wrappers." Bioinformatics (2012). DOI: 10.1093/bioinformatics/bts707

This project respects the original work and aims to provide a pure Python implementation for broader compatibility and ease of use in modern Python environments without requiring C compilation.

Features

• High Performance: Uses Numba to compile core algorithms to machine code, providing C-like performance.
• Pure Python compatible: Easy to install and modify.
• Algorithms:
  • MIC (Maximal Information Coefficient): APPROX-MIC and MIC_e estimators.
  • TIC (Total Information Coefficient)
  • GMIC (Generalized Mean Information Coefficient)
  • MAS (Maximum Asymmetry Score)
  • MEV (Maximum Edge Value)
  • MCN (Minimum Cell Number)

Installation

pip install .

Usage

import numpy as np
from fast_mine import mine_compute_score, MineParameter, MineProblem, mine_mic, EST_MIC_APPROX

# Generate data
x = np.linspace(0, 1, 100)
y = np.sin(10 * np.pi * x) + x

# Setup problem
prob = MineProblem(x=x, y=y, n=len(x))
param = MineParameter(alpha=0.6, c=15, est=EST_MIC_APPROX)

# Compute
score = mine_compute_score(prob, param)

# Get Statistics
print(f"MIC: {mine_mic(score)}")

Attribution & Acknowledgements

Example Datasets

Testing and verification for this library utilize datasets from the example-causal-datasets repository (CC0 1.0 Universal License).

We acknowledge the authors and contributors of the example-causal-datasets project for providing standardized real-world and synthetic datasets essential for benchmarking causal discovery and statistical association algorithms.

License

Based on minepy, this project is licensed under the GPL-3.0 License. See the LICENSE file for details.