Multivariate Maximal Correlation Analysis

Hoang Vu Nguyen, Emmanuel Müller, Jilles Vreeken, Pavel Efros, Klemens Böhm
; Proceedings of the 31st International Conference on Machine Learning, PMLR 32(2):775-783, 2014.

Abstract

Correlation analysis is one of the key elements of statistics, and has various applications in data analysis. Whereas most existing measures can only detect pairwise correlations between two dimensions, modern analysis aims at detecting correlations in multi-dimensional spaces. We propose MAC, a novel multivariate correlation measure designed for discovering multi-dimensional patterns. It belongs to the powerful class of maximal correlation analysis, for which we propose a generalization to multivariate domains. We highlight the limitations of current methods in this class, and address these with MAC. Our experiments show that MAC outperforms existing solutions, is robust to noise, and discovers interesting and useful patterns.

Cite this Paper


BibTeX
@InProceedings{pmlr-v32-nguyenc14, title = {Multivariate Maximal Correlation Analysis}, author = {Hoang Vu Nguyen and Emmanuel Müller and Jilles Vreeken and Pavel Efros and Klemens Böhm}, booktitle = {Proceedings of the 31st International Conference on Machine Learning}, pages = {775--783}, year = {2014}, editor = {Eric P. Xing and Tony Jebara}, volume = {32}, number = {2}, series = {Proceedings of Machine Learning Research}, address = {Bejing, China}, month = {22--24 Jun}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v32/nguyenc14.pdf}, url = {http://proceedings.mlr.press/v32/nguyenc14.html}, abstract = {Correlation analysis is one of the key elements of statistics, and has various applications in data analysis. Whereas most existing measures can only detect pairwise correlations between two dimensions, modern analysis aims at detecting correlations in multi-dimensional spaces. We propose MAC, a novel multivariate correlation measure designed for discovering multi-dimensional patterns. It belongs to the powerful class of maximal correlation analysis, for which we propose a generalization to multivariate domains. We highlight the limitations of current methods in this class, and address these with MAC. Our experiments show that MAC outperforms existing solutions, is robust to noise, and discovers interesting and useful patterns.} }
Endnote
%0 Conference Paper %T Multivariate Maximal Correlation Analysis %A Hoang Vu Nguyen %A Emmanuel Müller %A Jilles Vreeken %A Pavel Efros %A Klemens Böhm %B Proceedings of the 31st International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2014 %E Eric P. Xing %E Tony Jebara %F pmlr-v32-nguyenc14 %I PMLR %J Proceedings of Machine Learning Research %P 775--783 %U http://proceedings.mlr.press %V 32 %N 2 %W PMLR %X Correlation analysis is one of the key elements of statistics, and has various applications in data analysis. Whereas most existing measures can only detect pairwise correlations between two dimensions, modern analysis aims at detecting correlations in multi-dimensional spaces. We propose MAC, a novel multivariate correlation measure designed for discovering multi-dimensional patterns. It belongs to the powerful class of maximal correlation analysis, for which we propose a generalization to multivariate domains. We highlight the limitations of current methods in this class, and address these with MAC. Our experiments show that MAC outperforms existing solutions, is robust to noise, and discovers interesting and useful patterns.
RIS
TY - CPAPER TI - Multivariate Maximal Correlation Analysis AU - Hoang Vu Nguyen AU - Emmanuel Müller AU - Jilles Vreeken AU - Pavel Efros AU - Klemens Böhm BT - Proceedings of the 31st International Conference on Machine Learning PY - 2014/01/27 DA - 2014/01/27 ED - Eric P. Xing ED - Tony Jebara ID - pmlr-v32-nguyenc14 PB - PMLR SP - 775 DP - PMLR EP - 783 L1 - http://proceedings.mlr.press/v32/nguyenc14.pdf UR - http://proceedings.mlr.press/v32/nguyenc14.html AB - Correlation analysis is one of the key elements of statistics, and has various applications in data analysis. Whereas most existing measures can only detect pairwise correlations between two dimensions, modern analysis aims at detecting correlations in multi-dimensional spaces. We propose MAC, a novel multivariate correlation measure designed for discovering multi-dimensional patterns. It belongs to the powerful class of maximal correlation analysis, for which we propose a generalization to multivariate domains. We highlight the limitations of current methods in this class, and address these with MAC. Our experiments show that MAC outperforms existing solutions, is robust to noise, and discovers interesting and useful patterns. ER -
APA
Nguyen, H.V., Müller, E., Vreeken, J., Efros, P. & Böhm, K.. (2014). Multivariate Maximal Correlation Analysis. Proceedings of the 31st International Conference on Machine Learning, in PMLR 32(2):775-783

Related Material