Nyström M-Hilbert-Schmidt independence criterion

Kalinke, Florian and Szabo, Zoltan ORCID: 0000-0001-6183-7603 (2023) Nyström M-Hilbert-Schmidt independence criterion. Proceedings of Machine Learning Research, 216. pp. 1005-1015. ISSN 2640-3498

	Text (Nyström M-Hilbert-Schmidt Independence Criterion) - Accepted Version Download (460kB)
	Text (Nyström M-Hilbert-Schmidt Independence Criterion (Supplementary Material)) - Accepted Version Download (241kB)
	Text (kalinke23a-supplement) - Published Version Available under License ["licenses_description_unspecified" not defined]. Download (241kB)
	Text (kalinke23a) - Published Version Available under License ["licenses_description_unspecified" not defined]. Download (498kB)

• Author
• Publisher
• Organisation

Abstract

Kernel techniques are among the most popular and powerful approaches of data science. Among the key features that make kernels ubiquitous are (i) the number of domains they have been designed for, (ii) the Hilbert structure of the function class associated to kernels facilitating their statistical analysis, and (iii) their ability to represent probability distributions without loss of information. These properties give rise to the immense success of Hilbert-Schmidt independence criterion (HSIC) which is able to capture joint independence of random variables under mild conditions, and permits closed-form estimators with quadratic computational complexity (w.r.t. the sample size). In order to alleviate the quadratic computational bottleneck in large-scale applications, multiple HSIC approximations have been proposed, however these estimators are restricted to M=2 random variables, do not extend naturally to the M≥2 case, and lack theoretical guarantees. In this work, we propose an alternative Nyström-based HSIC estimator which handles the M≥2 case, prove its consistency, and demonstrate its applicability in multiple contexts, including synthetic examples, dependency testing of media annotations, and causal discovery.

Item Type:	Article
Additional Information:	© 2023 The Author.
Divisions:	Statistics
Subjects:	H Social Sciences > HA Statistics
Date Deposited:	22 Feb 2023 15:51
Last Modified:	20 Dec 2024 00:57
URI:	http://eprints.lse.ac.uk/id/eprint/118251

Actions (login required)

View Item