Presentation | 2011-12-20 Eigen Vector Descent and Line Search for Multilayer Perceptron Seiya SATOH, Ryohei NAKANO, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | As learning methods of a multilayer perceptron (MLP), we have the BP algorithm, Newton's method, quasi-Newton method, and so on. However, since the MLP search space is full of crevasse-like forms having huge condition numbers, it is very unlikely for such usual existing methods to perform efficient search in the space. This paper proposes a new search method which utilizes eigen vector descent and line search, stably finding excellent solutions in such an extraordinary search space. The proposed method is evaluated with promising results through experiments for MLPs having sigmoidal or exponential activation functions. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | multilayer perceptron / polynomial network / singular region / search method / line search |
Paper # | NC2011-87 |
Date of Issue |
Conference Information | |
Committee | NC |
---|---|
Conference Date | 2011/12/13(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 | Neurocomputing (NC) |
---|---|
Language | JPN |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | Eigen Vector Descent and Line Search for Multilayer Perceptron |
Sub Title (in English) | |
Keyword(1) | multilayer perceptron |
Keyword(2) | polynomial network |
Keyword(3) | singular region |
Keyword(4) | search method |
Keyword(5) | line search |
1st Author's Name | Seiya SATOH |
1st Author's Affiliation | Chubu University() |
2nd Author's Name | Ryohei NAKANO |
2nd Author's Affiliation | Chubu University |
Date | 2011-12-20 |
Paper # | NC2011-87 |
Volume (vol) | vol.111 |
Number (no) | 368 |
Page | pp.pp.- |
#Pages | 6 |
Date of Issue |