Large-Scale Sparse Kernel Canonical Correlation Analysis

The matlab folder contains the MATLAB codes applied in [1]. The python folder contains the python version of gradKCCA together with examples to simulate data.

Real Datasets

• MNIST Handwritten Digits
• MediaMill

Authors and Contact Information

• Viivi Uurtio * - [email protected]
• Sahely Bhadra * - [email protected]
• Juho Rousu - [email protected]

* Answer considerations regarding the codes

Reference

Viivi Uurtio, Sahely Bhadra, and Juho Rousu. Large-scale sparse kernel canonical correlation analysis. In Kamalika Chaudhuri and Ruslan Salakhutdinov, editors, Proceedings of the 36th International Conference on Machine Learning, volume 97 of Proceedings of Machine Learning Research, pages 6383–6391, Long Beach, California, USA, 09–15 Jun 2019. PMLR.

Update on February 2024

Fixed a few small bugs:

• Matlab: added centering after kernel evaluations
• Python:
  • Added centering after kernel evaluations
  • Fixed an indexing error in L1 norm
  • Fixed RBF kernel calculation
  • Introduced a more flexible way of defining kernel parameters using dictionaries for the parameters