On extensions of LARS by information geometry: Convex objectives and ̀p-norm

Masahiro Yukawa, Shun Ichi Amari

Research output: Contribution to conferencePaper

Abstract

This paper addresses extensions of the Least Angle Regression (LARS) algorithm from two different aspects: (i) from quadratic to more general objectives, and (ii) from ̀1-norm to ̀p-norm for p < 1. The equiangular vector, which is the key of LARS, is reproduced in connection with the Riemannian metric induced by the objective function, thereby making the extensions feasible. It is shown, in the case of p < 1, that two types of trajectory . the c-trajectory and the λ-trajectory . need to be distinguished by revealing the discontinuity of the λ-trajectory.

Original languageEnglish
Pages326-331
Number of pages6
Publication statusPublished - 2011 Dec 1
EventAsia-Pacific Signal and Information Processing Association Annual Summit and Conference 2011, APSIPA ASC 2011 - Xi'an, China
Duration: 2011 Oct 182011 Oct 21

Other

OtherAsia-Pacific Signal and Information Processing Association Annual Summit and Conference 2011, APSIPA ASC 2011
CountryChina
CityXi'an
Period11/10/1811/10/21

    Fingerprint

ASJC Scopus subject areas

  • Information Systems
  • Signal Processing

Cite this

Yukawa, M., & Amari, S. I. (2011). On extensions of LARS by information geometry: Convex objectives and ̀p-norm. 326-331. Paper presented at Asia-Pacific Signal and Information Processing Association Annual Summit and Conference 2011, APSIPA ASC 2011, Xi'an, China.