loading...
-EM Algorithm and -ICA Learning Based upon Extended Logarithmic Information Measures
Como, Italy July 24-July 27
DOI Bookmark: http://doi.ieeecomputersociety.org/10.1109/IJCNN.2000.861329IEEE-INNS-ENNS International Joint Co ...
 This Article 
 
Buy PURCHASE ARTICLE: $19
PDF
HTML
Buy IEEE Xplore Subscribers
 
 Share 
   
 Bibliographic References 
   
 Add to: 
 
Digg
Furl
Spurl
Blink
Simpy
Google
Del.icio.us
Y!MyWeb
 
 Search 
   
ASCII Text
x 
 
Yasuo Matsuyama, Takeshi Niimoto, Naoto Katsumata, Yoshitaka Suzuki, Satoshi Furukawa, "-EM Algorithm and -ICA Learning Based upon Extended Logarithmic Information Measures," Neural Networks, IEEE - INNS - ENNS International Joint Conference on, vol. 3, pp. 3351, IEEE-INNS-ENNS International Joint Conference on Neural Networks (IJCNN'00)-Volume 3, 2000.
 
 
BibTex
x
 
@article{ 10.1109/IJCNN.2000.861329,
author = {Yasuo Matsuyama and Takeshi Niimoto and Naoto Katsumata and Yoshitaka Suzuki and Satoshi Furukawa},
title = {-EM Algorithm and -ICA Learning Based upon Extended Logarithmic Information Measures},
journal ={Neural Networks, IEEE - INNS - ENNS International Joint Conference on},
volume = {3},
year = {2000},
issn = {1098-7576},
pages = {3351},
doi = {http://doi.ieeecomputersociety.org/10.1109/IJCNN.2000.861329},
publisher = {IEEE Computer Society},
address = {Los Alamitos, CA, USA},
}
 
 
RefWorks Procite/RefMan/Endnote
x 
 
TY - CONF
JO - Neural Networks, IEEE - INNS - ENNS International Joint Conference on
TI - -EM Algorithm and -ICA Learning Based upon Extended Logarithmic Information Measures
SN - 1098-7576
SP
EP
A1 - Yasuo Matsuyama,
A1 - Takeshi Niimoto,
A1 - Naoto Katsumata,
A1 - Yoshitaka Suzuki,
A1 - Satoshi Furukawa,
PY - 2000
VL - 3
JA - Neural Networks, IEEE - INNS - ENNS International Joint Conference on
ER -
 
Yasuo Matsuyama, Waseda University
Takeshi Niimoto, Waseda University
Naoto Katsumata, Waseda University
Yoshitaka Suzuki, Waseda University
Satoshi Furukawa, Waseda University
The \math-logarithm extends the logarithm as the special case of \math = - 1. Usage of \math-related information measures based upon this extended logarithm is expected to be effective to speedup of convergence, i.e., on the improvement of learning aptitude. In this paper, two typical cases are investigated. One is the \math-EM algorithm (\math-Expectation-Maximization algorithm), which is derived from the \math-log-likelihood ratio. The other is the \math-ICA (\math-Independent Component Analysis), which is formulated as minimizing the \math-mutual information. In the derivation of both algorithms, the \math-divergence plays the main role. For the \math-EM algorithm, the reason for the speedup is explained using Hessian and Jacobian matrices for learning. For the \math-ICA learning, methods of exploiting the past and future information are presented. Examples are shown on single-loop \math-EM's and sample-based \math-ICA's. In all cases, effective speedups are observed. Thus, this paper's examples together with formerly reported ones are evidences that the speed improvement by the \math-logarithm is a general property beyond individual problems.
Citation:
Yasuo Matsuyama, Takeshi Niimoto, Naoto Katsumata, Yoshitaka Suzuki, Satoshi Furukawa, "-EM Algorithm and -ICA Learning Based upon Extended Logarithmic Information Measures," ijcnn, vol. 3, pp.3351, IEEE-INNS-ENNS International Joint Conference on Neural Networks (IJCNN'00)-Volume 3, 2000
Usage of this product signifies your acceptance of the Terms of Use.