Loading [a11y]/accessibility-menu.js
Yast Algorithm for Minor Subspace Tracking | IEEE Conference Publication | IEEE Xplore

Yast Algorithm for Minor Subspace Tracking


Abstract:

This paper introduces a new algorithm for tracking the minor subspace of the correlation matrix associated with time series. This algorithm is shown to have a better conv...Show More

Abstract:

This paper introduces a new algorithm for tracking the minor subspace of the correlation matrix associated with time series. This algorithm is shown to have a better convergence rate than existing methods. Moreover, it guarantees the orthonormality of the subspace weighting matrix at each iteration, and reaches a linear complexity
Date of Conference: 14-19 May 2006
Date Added to IEEE Xplore: 24 July 2006
Print ISBN:1-4244-0469-X

ISSN Information:

Conference Location: Toulouse, France
Département TSI, Télécom Paris Technology, Paris, France
Département TSI, Télécom Paris Technology, Paris, France
Département TSI, Télécom Paris Technology, Paris, France

1. INTRODUCTION

Fast estimation and tracking of the principal or minor subspace of a sequence of random vectors is a major problem in many applications. We can cite, for example, code division multiple access (CDMA) communications, where many multiuser detection algorithms are actually subspace-based [1]. Recently, we presented in [2] a new principal subspace tracker dedicated to time series analysis, which is derived from the SP algorithm by C.E. Davila [3]. This new algorithm, referred to as YAST, reaches the lowest complexity found in the lit-erature, and outperforms classical methods in terms of subspace esti-mation. Moreover, it guarantees the orthonormality of the subspace weighting matrix at each time step. In this paper, we focus on minor subspace analysis (MSA). In the literature, it is commonly admitted that MSA is a more difficult problem than principal subspace analysis (PCA). In particular, the classical Oja algorithm [4] is known to diverge. Some more robust MSA algorithms have been presented in [5]–[9]. However the convergence rate of these algorithms is much lower than that of the classical PCA techniques. Here we propose a version of the YAST algorithm dedicated to MSA, which is shown to have better convergence properties.

Département TSI, Télécom Paris Technology, Paris, France
Département TSI, Télécom Paris Technology, Paris, France
Département TSI, Télécom Paris Technology, Paris, France

References

References is not available for this document.