collasso 0.1.0

Sparse Linear Multi-Task Regression

Sparse Linear Multi-Task Regression

Scope

The Python package collasso implements sparse linear multi-task regression with correlation-based information sharing (Rauschenberger, 2026). In contrast to MultiTaskLassoCV from scikit-learn, it supports target-specific feature selection, target-specific feature matrices, and privileged information.

Installation

Install the latest release from PyPI or Anaconda (not yet available):

pip install -U collasso
conda install -c conda-forge collasso

Alternatively, install the development version from GitHub or TestPyPI:

pip install -U git+https://github.com/rauschenberger/collasso.git
pip install -i https://test.pypi.org/simple/ collasso

Usage

Use the class CoopLassoCV to model a multivariate target (n × q matrix Y) based on high-dimensional features (n × p matrix or n × p × q array X), potentially indicating primary and auxiliary features (p-dimensional vector or p × q matrix Z).

from collasso import CoopLassoCV
model = CoopLassoCV()
model.fit(X_train, y_train, Z)
model.coef_ # estimated coefficients
model.predict(X_test) # out-of-sample predictions

Please find the full documentation on the website. The vignette contains examples on multi-task regression with a common feature matrix, multi-task regression with specific feature matrices, and multi-task regression with privileged information.

Reference

Armin Rauschenberger (2026). "Sparse linear multi-task regression with correlation-based information sharing". Manuscript in preparation.

Repository

The source code of this Python package is available on GitHub. This personal GitHub repository is mirrored at two institutional GitLab instances (see LIH and LCSB).

Disclosure

Large-language models (mainly Claude Sonnet 4.6 and Claude Opus 4.6) were used for reviewing Python code and documentation as well as for drafting or reviewing configuration files (.toml and .yaml).

Disclaimer

Copyright © 2026 Armin Rauschenberger; Luxembourg Institute of Health (LIH), Department of Medical Informatics (DMI), Bioinformatics and Artificial Intelligence (BioAI); University of Luxembourg, Luxembourg Centre for Systems Biomedicine (LCSB), Biomedical Data Science (BDS). This Python package is distributed under the BSD-3-Clause license.