Presentation 2012-11-08
Ancestral Atom Learning for Dictionary Generation in Sparse Coding
Toshimitsu ARITAKE, Hideitsu HINO, Noboru MURATA,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) Sparse Coding is a methodology to represent signals with combinations of only a small number of basis vectors. In sparse coding, designing dictionary is a fundamental problem. An approach for desigining dictionary which adapts observed signals is learning from observed signals. In this paper, like wavelet analysis, a dictionary for sparse signal representation is assumed to be generated from single vector called ancestral atom, and a method for learning the ancestral atom is proposed. Experimental results of ancestral atom learning by proposed method with both artificial and real-world seismic signal are shown to exhibit characteristics and advantages of the proposed algorithm.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) atom decomposition / sparse representation / structured dictionary learning
Paper # IBISML2012-82
Date of Issue

Conference Information
Committee IBISML
Conference Date 2012/10/31(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Information-Based Induction Sciences and Machine Learning (IBISML)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Ancestral Atom Learning for Dictionary Generation in Sparse Coding
Sub Title (in English)
Keyword(1) atom decomposition
Keyword(2) sparse representation
Keyword(3) structured dictionary learning
1st Author's Name Toshimitsu ARITAKE
1st Author's Affiliation Waseda University()
2nd Author's Name Hideitsu HINO
2nd Author's Affiliation Waseda University
3rd Author's Name Noboru MURATA
3rd Author's Affiliation Waseda University
Date 2012-11-08
Paper # IBISML2012-82
Volume (vol) vol.112
Number (no) 279
Page pp.pp.-
#Pages 8
Date of Issue