icspylab 1.0.0

Invariant Coordinate Selection (ICS) for multivariate data analysis.

ICSpyLab

Overview

Invariant Coordinate Selection (ICS) is a data transformation method. It transforms the data, via the simultaneous diagonalization of two scatter matrices, into an invariant coordinate system or independent components, depending on the underlying assumptions. It is particularly useful for dimension reduction. Unlike PCA, ICS is not based on variance maximization but on the maximization/minimization of a generalized kurtosis, and it is invariant not only to orthogonal data transformations but to any affine transformation.

This package brings the main functionalities of the ICS R package to Python, offering tools for identifying and selecting invariant coordinates in multivariate data. It includes various covariance estimators, transformation settings, and plotting utilities. Our extensive testing ensures results consistent with the R package, making it easy for users to transition from R to Python or start fresh with ICS.

Check out the documentation for more details.

Installation

pip install numpy pandas scipy scikit-learn seaborn matplotlib 
pip install icspylab

Usage

from icspylab import ICS, cov, covW
from sklearn.datasets import load_iris

# Load dataset
iris = load_iris()
X = iris.data

# Instantiate ICS object
# ics = ICS() # default parameters
ics = ICS(S1=cov, S2=covW, algorithm='standard', S2_args={'alpha': 1, 'cf': 2})

# Fit and transform the ICS model (equivalent of the function ICS-S3() from the R package ICS)
ics.fit_transform(X)

# Printing a summary
ics.describe()

🤝 Contributing

We welcome contributions of all kinds, including bug fixes, new features, and documentation improvements.

To get started, please check out our CONTRIBUTING.md guide.

Citation

If you use this software, please cite:

Becquart, C. and Abdelsameia, A. (2026). ICSpyLab (Version 1.0.0) https://doi.org/10.5281/zenodo.20310665

@software{becquart2026,
 author = {Becquart, Colombe and Abdelsameia, Abdallah},
 title = {ICSpyLab},
 year = {2026},
 version = {1.0.0},
 doi = {10.5281/zenodo.20310665}
}