This paper introduces a novel mathematical and computational framework, namely Log-Hilbert-Schmidt metric between positive definite operators on a Hilbert space. This is a generalization of the Log-Euclidean metric on the Rie-mannian manifold of positive definite matrices to the infinite-dimensional setting. The general framework is applied in particular to compute distances between co-variance operators on a Reproducing Kernel Hilbert Space (RKHS), for which we obtain explicit formulas via the corresponding Gram matrices. Empirically, we apply our formulation to the task of multi-category image classification, where each image is represented by an infinite-dimensional RKHS covariance operator. On several challenging datasets, our method significantly outperforms approaches based on covariance matrices computed directly on the original input features, including those using the Log-Euclidean metric, Stein and Jeffreys divergences, achieving new state of the art results.

Log-Hilbert-Schmidt metric between positive definite operators on Hilbert spaces

MURINO, Vittorio
2014-01-01

Abstract

This paper introduces a novel mathematical and computational framework, namely Log-Hilbert-Schmidt metric between positive definite operators on a Hilbert space. This is a generalization of the Log-Euclidean metric on the Rie-mannian manifold of positive definite matrices to the infinite-dimensional setting. The general framework is applied in particular to compute distances between co-variance operators on a Reproducing Kernel Hilbert Space (RKHS), for which we obtain explicit formulas via the corresponding Gram matrices. Empirically, we apply our formulation to the task of multi-category image classification, where each image is represented by an infinite-dimensional RKHS covariance operator. On several challenging datasets, our method significantly outperforms approaches based on covariance matrices computed directly on the original input features, including those using the Log-Euclidean metric, Stein and Jeffreys divergences, achieving new state of the art results.
2014
Engineering controlled terms: Covariance matrix; Hilbert spaces; Information science; Mathematical operators; Vector spaces Computational framework; Covariance matrices; Covariance operators; Explicit formula; Infinite dimensional; Positive definite; Positive-definite matrices; Reproducing Kernel Hilbert spaces Engineering main heading: Image classification
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11562/961554
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 56
  • ???jsp.display-item.citation.isi??? ND
social impact