Presentation | 2018-03-06 Natural gradient method by the symmetry in data for complex valued neural networks Junichi Mukuno, Hajime Matsui, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | On the real valued neural networks, the natural gradient method is proposed by Amari to resolve the problem the speed of learning is slow by using Riemannian metric and moreover is known that there is good properties. We consider a method extending the natural gradient method to complex valued neural networks in this research. Complex numbers are suited for representing reflections and rotations in the plane. By focusing on its property, we show that the output function of the complex neural networks can have a property of conjugate invariance and rotation invariance when data set is conjugate invariant and rotation invariant, respectively. Through simulation experiences, we check the performance of our proposal. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | complex valued neural network / natural gradient / symmetry of data |
Paper # | IBISML2017-104 |
Date of Issue | 2018-02-26 (IBISML) |
Conference Information | |
Committee | IBISML |
---|---|
Conference Date | 2018/3/5(2days) |
Place (in Japanese) | (See Japanese page) |
Place (in English) | Nishijin Plaza, Kyushu University |
Topics (in Japanese) | (See Japanese page) |
Topics (in English) | Statisitical Mathematics, Machine Learning, Data Mining, etc. |
Chair | Kenji Fukumizu(ISM) |
Vice Chair | Masashi Sugiyama(Univ. of Tokyo) / Hisashi Kashima(Kyoto Univ.) |
Secretary | Masashi Sugiyama(Nagoya Inst. of Tech.) / Hisashi Kashima(Univ. of Tokyo) |
Assistant | Tomoharu Iwata(NTT) / Toshihiro Kamishima(AIST) |
Paper Information | |
Registration To | Technical Committee on Infomation-Based Induction Sciences and Machine Learning |
---|---|
Language | JPN |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | Natural gradient method by the symmetry in data for complex valued neural networks |
Sub Title (in English) | |
Keyword(1) | complex valued neural network |
Keyword(2) | natural gradient |
Keyword(3) | symmetry of data |
1st Author's Name | Junichi Mukuno |
1st Author's Affiliation | Toyota Technological Institute(Toyota Technological Inst.) |
2nd Author's Name | Hajime Matsui |
2nd Author's Affiliation | Toyota Technological Institute(Toyota Technological Inst.) |
Date | 2018-03-06 |
Paper # | IBISML2017-104 |
Volume (vol) | vol.117 |
Number (no) | IBISML-475 |
Page | pp.pp.97-102(IBISML), |
#Pages | 6 |
Date of Issue | 2018-02-26 (IBISML) |