Presentation 2019-03-08
Typical performance of the L1 regularization regression from linear measurements with measurement noise and large coherence
Minori Ihara, Kazunori Iwata, Kazushi Mimura,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) We evaluate typical performance of compressed sensing in the case where iterative recovery algorithms fail to converge. The simple example in this case is where elements of measurement matrix obey probability distribution with non-zero mean. In this study, we especially focus on the case of measurement matrix has large coherence, and evaluate the mean square error (MSE) between an original signal to be estimated and its estimate by using the replica method.When coherence of measurement matrix is enough large, the performance depends only on signal density.And the MSE sometimes behaves non-monotonically as compression rate increases under the influence of measurement noise.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) replica method / compressed sensing / L1 regularization regression / coherence
Paper # IT2018-117,ISEC2018-123,WBS2018-118
Date of Issue 2019-02-28 (IT, ISEC, WBS)

Conference Information
Committee IT / ISEC / WBS
Conference Date 2019/3/7(2days)
Place (in Japanese) (See Japanese page)
Place (in English) University of Electro-Communications
Topics (in Japanese) (See Japanese page)
Topics (in English) joint meeting of IT, ISEC, and WBS
Chair Jun Muramatsu(NTT) / Atsushi Fujioka(Kanagawa Univ.) / Minoru Okada(NAIST)
Vice Chair Tadashi Wadayama(Nagoya Inst. of Tech.) / Shiho Moriai(NICT) / Shoichi Hirose(Univ. of Fukui) / Koji Ohuchi(Shizuoka Univ.) / Kenichi Takizawa(NICT)
Secretary Tadashi Wadayama(Nagano Pref Inst. of Tech.) / Shiho Moriai(UEC) / Shoichi Hirose(Tokai Univ.) / Koji Ohuchi(NICT) / Kenichi Takizawa(Ibaraki Univ.)
Assistant Takahiro Yoshida(Yokohama College of Commerce) / Kazunari Omote(Tsukuba Univ.) / Yuuji Suga(IIJ) / Ryohei Nakamura(National Defense Academy) / Duong Quang Thang(NAIST)

Paper Information
Registration To Technical Committee on Information Theory / Technical Committee on Information Security / Technical Committee on Wideband System
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Typical performance of the L1 regularization regression from linear measurements with measurement noise and large coherence
Sub Title (in English)
Keyword(1) replica method
Keyword(2) compressed sensing
Keyword(3) L1 regularization regression
Keyword(4) coherence
1st Author's Name Minori Ihara
1st Author's Affiliation Hiroshima City University(Hiroshima City Univ.)
2nd Author's Name Kazunori Iwata
2nd Author's Affiliation Hiroshima City University(Hiroshima City Univ.)
3rd Author's Name Kazushi Mimura
3rd Author's Affiliation Hiroshima City University(Hiroshima City Univ.)
Date 2019-03-08
Paper # IT2018-117,ISEC2018-123,WBS2018-118
Volume (vol) vol.118
Number (no) IT-477,ISEC-478,WBS-479
Page pp.pp.257-262(IT), pp.257-262(ISEC), pp.257-262(WBS),
#Pages 6
Date of Issue 2019-02-28 (IT, ISEC, WBS)