Presentation | 1999/5/20 A Neural Network with Trainable Nonlinear Connections and Activation Functions and Its Learning Algorithm Issei IDO, Kenji NAKAYAMA, Akihiro HIRANO, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | It is important to realize small size neural networks by which desired functions are achieved. Regarding the activation functions, a method was proposed, in which an abtivation function is composed of several basic functions. The unit input is given as a linear combination of the inputs. The unit output takes the same value for this linear combination. In this paper, a nonlinear function is used to express the unit input in order to expand a degree of freedom. The nonlinear function parameters and the above activation functions are simultaneously trained. More complex classification problems can be solved using a small size neural network. Fast and stable convergence and small size neural networks are confirmed through computer simulation using several classification problems. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | Neural networks / Activation function / Nonlinear function / Simultaneous learning / Classification |
Paper # | NC99-8 |
Date of Issue |
Conference Information | |
Committee | NC |
---|---|
Conference Date | 1999/5/20(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) | A Neural Network with Trainable Nonlinear Connections and Activation Functions and Its Learning Algorithm |
Sub Title (in English) | |
Keyword(1) | Neural networks |
Keyword(2) | Activation function |
Keyword(3) | Nonlinear function |
Keyword(4) | Simultaneous learning |
Keyword(5) | Classification |
1st Author's Name | Issei IDO |
1st Author's Affiliation | Div. of Electronics and Computer Science, Master's Level Section Graduate School of Natural Science and Technology, Kanazawa University() |
2nd Author's Name | Kenji NAKAYAMA |
2nd Author's Affiliation | Div. of Electronics and Computer Science, Master's Level Section Graduate School of Natural Science and Technology, Kanazawa University |
3rd Author's Name | Akihiro HIRANO |
3rd Author's Affiliation | Div. of Electronics and Computer Science, Master's Level Section Graduate School of Natural Science and Technology, Kanazawa University |
Date | 1999/5/20 |
Paper # | NC99-8 |
Volume (vol) | vol.99 |
Number (no) | 58 |
Page | pp.pp.- |
#Pages | 8 |
Date of Issue |