Presentation 2002/3/12
Meta Learning for Quick Incremental Learning
Takayuki OHIRA, Koichiro YAMAUCHI, Takashi OMORI,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) The neural networks need a high technique to deal with incremental learning. This is because one parameter of the system plays to record not only one instance but also several different instances. To overcome the problem, several systems, which have a buffer beside of the neural network, have been proposed. The buffer stores some of old presented instances. The neural network learns a new instance with the stored instances in the buffer. However, with this technique, the learner wastes a long time to learn the instances in the buffer. To overcome the drawback, we propose here a novel technique to use a meta-learning network, which learns a strategy for the incremental learning.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) quick incremental learning / meta-learning / learning to learn / learning strategy / RBF / RAN
Paper # NC2001-178
Date of Issue

Conference Information
Committee NC
Conference Date 2002/3/12(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) Meta Learning for Quick Incremental Learning
Sub Title (in English)
Keyword(1) quick incremental learning
Keyword(2) meta-learning
Keyword(3) learning to learn
Keyword(4) learning strategy
Keyword(5) RBF
Keyword(6) RAN
1st Author's Name Takayuki OHIRA
1st Author's Affiliation Faculty of Engineering, Hokkaido University()
2nd Author's Name Koichiro YAMAUCHI
2nd Author's Affiliation Faculty of Engineering, Hokkaido University
3rd Author's Name Takashi OMORI
3rd Author's Affiliation Faculty of Engineering, Hokkaido University
Date 2002/3/12
Paper # NC2001-178
Volume (vol) vol.101
Number (no) 736
Page pp.pp.-
#Pages 7
Date of Issue