Presentation | 2002/12/6 A Proposal of The Neural Network Using Hints on Conditional Quantiles to Improve Its Generalization Performance Noriyuki YAMANAKA, Ichiro TAKEUCHI, Takeshi FURUHASHI, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | We propose, in this paper, a new approach to improve the generalization performance of artificial neural networks (NN) for regression problems. The basic idea is to incorpolate a robust statistics technique into NN within the framework of learning with hints. In particular, we propose to train NN with the hints on conditional quantiles, and show that this approach yield a robust estimation of conditional expectation, which we want to estimate in regression problem, and improve its generalization property. We provide, in this paper, an numerical study on our proposal to show its effectiveness and characteristics as well as the comparison with conventional Weight Decay technique, which is one of the most common technique for generalization. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | Neural Network / L_1 Regression / Robust Statistics / Learning with Hints / Quantile Regression / Generalization Performance / Over Training |
Paper # | NC2002-92 |
Date of Issue |
Conference Information | |
Committee | NC |
---|---|
Conference Date | 2002/12/6(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 Proposal of The Neural Network Using Hints on Conditional Quantiles to Improve Its Generalization Performance |
Sub Title (in English) | |
Keyword(1) | Neural Network |
Keyword(2) | L_1 Regression |
Keyword(3) | Robust Statistics |
Keyword(4) | Learning with Hints |
Keyword(5) | Quantile Regression |
Keyword(6) | Generalization Performance |
Keyword(7) | Over Training |
1st Author's Name | Noriyuki YAMANAKA |
1st Author's Affiliation | Faculty of Engineering, Mie University() |
2nd Author's Name | Ichiro TAKEUCHI |
2nd Author's Affiliation | Faculty of Engineering, Mie University |
3rd Author's Name | Takeshi FURUHASHI |
3rd Author's Affiliation | Faculty of Engineering, Mie University |
Date | 2002/12/6 |
Paper # | NC2002-92 |
Volume (vol) | vol.102 |
Number (no) | 508 |
Page | pp.pp.- |
#Pages | 6 |
Date of Issue |