Presentation 2020-08-28
Improvement Convergence Rate of the Sign Algorithm by Natural Gradient Method
Taiyo Mineo, Hayaru Shouno,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) In lossless audio compression, it is essential to predictive residuals to be sparse, since we apply entropy codings to residuals. The Sign Algorithm(SA) is the conventional method to minimize the residuals in magnitude, but it has poor convergence performance than the Least Mean Square(LMS) Algorithm. In this paper, we discuss the improvement of the convergence performance of the SA by the natural gradient method. We show that the auto-correlation matrix of the input signal is needed to apply the natural gradient. The variable step-size algorithm has some relations to well-known adaptive algorithms such as the NLMS and RLS. We also show that the proposed methods are better convergence performance than the SA for toy data and useful music data through computer experiences.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Lossless Audio Coding / Adaptive Algorithm / Sign Algorithm / Natural Gradient Method
Paper # SIP2020-34
Date of Issue 2020-08-20 (SIP)

Conference Information
Committee SIP
Conference Date 2020/8/27(2days)
Place (in Japanese) (See Japanese page)
Place (in English) Online
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair Kazunori Hayashi(Kyoto Univ.)
Vice Chair Yukihiro Bandou(NTT) / Toshihisa Tanaka(Tokyo Univ. Agri.&Tech.)
Secretary Yukihiro Bandou(Hosei Univ.) / Toshihisa Tanaka(Waseda Univ.)
Assistant Yuichi Tanaka(Tokyo Univ. Agri.&Tech.)

Paper Information
Registration To Technical Committee on Signal Processing
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Improvement Convergence Rate of the Sign Algorithm by Natural Gradient Method
Sub Title (in English)
Keyword(1) Lossless Audio Coding
Keyword(2) Adaptive Algorithm
Keyword(3) Sign Algorithm
Keyword(4) Natural Gradient Method
1st Author's Name Taiyo Mineo
1st Author's Affiliation The University of Electro-Communications(UEC)
2nd Author's Name Hayaru Shouno
2nd Author's Affiliation The University of Electro-Communications(UEC)
Date 2020-08-28
Paper # SIP2020-34
Volume (vol) vol.120
Number (no) SIP-142
Page pp.pp.19-24(SIP),
#Pages 6
Date of Issue 2020-08-20 (SIP)