Presentation 2018-03-06
Applicability of Fast Decimation Algorithm
Daisuke Motoki, Shohei Watabe, Tetsuro Nikuni,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) A decimation algorithm was developed by Decelle et al. for an inverse problem optimization method, which sequentially reduces the number of parameters for each epoch of machine learning.This algorithm has the merit that arbitrariness of parameters of L1 regularization method which is often used in sparse modeling can be removed. However, the decimation algorithm has problems in terms of calculation accuracy and speed.In order to improve the problems raised above, we developed Fast Decimation Algorithm that does not use maximum likelihood in the early epochs of learning. We applied our algorithm to a two-parameter Boltzmann Machine applicable to Item Responsive Theory with relevant structure between examinees and demonstrated its usefulness by making use of the Area Under the ROC Curve (AUC).
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Statistical Machine Learning / Sparse Modeling / Boltzmann Machine
Paper # IBISML2017-103
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) Applicability of Fast Decimation Algorithm
Sub Title (in English) Sparse two-parameter Boltzmann machine as a benchmark function
Keyword(1) Statistical Machine Learning
Keyword(2) Sparse Modeling
Keyword(3) Boltzmann Machine
1st Author's Name Daisuke Motoki
1st Author's Affiliation Tokyo University of Science(Tokyo Univ. of Science)
2nd Author's Name Shohei Watabe
2nd Author's Affiliation Tokyo University of Science(Tokyo Univ. of Science)
3rd Author's Name Tetsuro Nikuni
3rd Author's Affiliation Tokyo University of Science(Tokyo Univ. of Science)
Date 2018-03-06
Paper # IBISML2017-103
Volume (vol) vol.117
Number (no) IBISML-475
Page pp.pp.91-95(IBISML),
#Pages 5
Date of Issue 2018-02-26 (IBISML)