Presentation 2002/1/22
Neural Network Pruning using MV Regularizer
Takashi NAGATA, Atsushi KAWATA, Ken-ichi YAMADA, Ryohei NAKANO,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) The paper proposes a new method of neural network pruning by using a strong regularizer called MV regularizer. The MV regularizer learns a distinct penalty factor attached to each weight by solving a minimization problem over the validation error. After the learning, penalty factors are used to prune network weights one by one by monitoring generalization performance. The experiments using artificial data and real data showed the proposed method worked very well to remove a number of insignificant weights without the loss of generalization.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) neural network / network pruning / MV regularizer / generalization
Paper #
Date of Issue

Conference Information
Committee NC
Conference Date 2002/1/22(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) Neural Network Pruning using MV Regularizer
Sub Title (in English)
Keyword(1) neural network
Keyword(2) network pruning
Keyword(3) MV regularizer
Keyword(4) generalization
1st Author's Name Takashi NAGATA
1st Author's Affiliation Department of Intelligence and Computer Science, Nagoya Institute of Technology()
2nd Author's Name Atsushi KAWATA
2nd Author's Affiliation Department of Intelligence and Computer Science, Nagoya Institute of Technology
3rd Author's Name Ken-ichi YAMADA
3rd Author's Affiliation Department of Intelligence and Computer Science, Nagoya Institute of Technology
4th Author's Name Ryohei NAKANO
4th Author's Affiliation Department of Intelligence and Computer Science, Nagoya Institute of Technology
Date 2002/1/22
Paper #
Volume (vol) vol.101
Number (no) 616
Page pp.pp.-
#Pages 6
Date of Issue