Welcome to EffDim
EffDim is a unified, research-oriented Python library designed to compute "effective dimensionality" (ED) across diverse data modalities.
It aims to standardize the fragmented landscape of ED metrics found in statistics, physics, information theory, and machine learning into a single, cohesive interface.
Key Features
- Modality Agnostic: Works with raw data, covariance matrices, and pre-computed spectra.
- Unified Interface: Simple
computeandanalyzefunctions. - Extensive Estimators: PCA, Participation Ratio, Shannon Entropy, and more.
- Research Ready: Accurate implementations of metrics from literature.
Installation
Install via pip:
Quick Start
import numpy as np
import effdim
# Generate random high-dimensional data
data = np.random.randn(100, 50)
# Compute Effective Dimension using PCA (95% variance)
ed = effdim.compute(data, method='pca', threshold=0.95)
print(f"Effective Dimension (PCA): {ed}")
# Compute using Participation Ratio
pr = effdim.compute(data, method='participation_ratio')
print(f"Participation Ratio: {pr}")
Explore the User Guide for more examples.