aalto-ics-kepaco/scca-hsic

MATLAB implementation of SCCA-HSIC

Sparse Non-Linear CCA through Hilbert-Schmidt Independence Criterion

This repository contains the MATLAB codes applied in the experiments of [1].

Contents

• The scca-hsic folder contains the implementation of SCCA-HSIC and its Nyström approximated variant.

• The experiments folder contains scripts that can be used to analyse the performance of SCCA-HSIC when the number of related and noise variables increases. Scripts to test the scalability of the Nyström variant are also included.

• The tutorial folder contains example scripts to get acquainted with SCCA-HSIC.

Real Datasets

• Boston Housing Dataset
• Body Fat Dataset

Authors and Contact Information

• Viivi Uurtio * - viivi.uurtio@aalto.fi
• Sahely Bhadra * - sahely@iitpkd.ac.in
• Juho Rousu - juho.rousu@aalto.fi

* Answer considerations regarding the codes

Reference

[1] Uurtio, V., Bhadra, S., Rousu, J. Sparse Non-Linear CCA through Hilbert-Schmidt Independence Criterion. IEEE International Conference on Data Mining (ICDM 2018), to appear