Presentation | 2017-03-20 Selection of Near-Boundary Data for Semi-Supervised Learning Ryohei Tanaka, Xiao Ding, Soichiro Ono, Akio Furuhata, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | Semi-supervised learning (SSL) is a technique which makes use of unlabeled data in addition to labeled data to obtain better learning accuracies. The most fundamental and widely applicable subtype of SSL is called self-training, which produces additional labeled data from unlabeled data using the results of the classifier trained with the existing labeled data. Conventionally, the additional labeling is performed only on unlabeled data with high prediction confidence above the built-in threshold predetermined by the designer. On the other hand, it is known that learning data near the decision boundary play a crucial role for classification performances. In this paper, we introduce this knowledge to self-training by selectively labeling unconfident unlabeled data near the decision boundary. We also propose a novel method using optimal region accumulation which automatically optimizes the labeling threshold to accumulate data near the boundary. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | Semi-supervised learning / self-training / subspace method / handwritten digits recognition |
Paper # | BioX2016-33,PRMU2016-196 |
Date of Issue | 2017-03-13 (BioX, PRMU) |
Conference Information | |
Committee | PRMU / BioX |
---|---|
Conference Date | 2017/3/20(2days) |
Place (in Japanese) | (See Japanese page) |
Place (in English) | |
Topics (in Japanese) | (See Japanese page) |
Topics (in English) | |
Chair | Eisaku Maeda(NTT) / Masakatsu Nishigaki(Shizuoka Univ.) |
Vice Chair | Seiichi Uchida(Kyushu Univ.) / Hironobu Fujiyoshi(Chubu Univ.) / Akira Otsuka(AIST) / Hiroshi Takano(Toyama Pref. Univ.) |
Secretary | Seiichi Uchida(Kyoto Univ.) / Hironobu Fujiyoshi(NTT) / Akira Otsuka(NEC) / Hiroshi Takano(AIST) |
Assistant | Masaki Oonishi(AIST) / Takuya Funatomi(NAIST) / Masatsugu Ichino(Univ. of Electro-Comm.) / Naoyuki Takada(Secom) / Takahiro Aoki(Fujitsu Labs.) |
Paper Information | |
Registration To | Technical Committee on Pattern Recognition and Media Understanding / Technical Committee on Biometrics |
---|---|
Language | JPN |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | Selection of Near-Boundary Data for Semi-Supervised Learning |
Sub Title (in English) | |
Keyword(1) | Semi-supervised learning |
Keyword(2) | self-training |
Keyword(3) | subspace method |
Keyword(4) | handwritten digits recognition |
1st Author's Name | Ryohei Tanaka |
1st Author's Affiliation | Toshiba Corporation(Toshiba) |
2nd Author's Name | Xiao Ding |
2nd Author's Affiliation | Toshiba Corporation(Toshiba) |
3rd Author's Name | Soichiro Ono |
3rd Author's Affiliation | Toshiba Corporation(Toshiba) |
4th Author's Name | Akio Furuhata |
4th Author's Affiliation | Toshiba Corporation(Toshiba) |
Date | 2017-03-20 |
Paper # | BioX2016-33,PRMU2016-196 |
Volume (vol) | vol.116 |
Number (no) | BioX-527,PRMU-528 |
Page | pp.pp.1-6(BioX), pp.1-6(PRMU), |
#Pages | 6 |
Date of Issue | 2017-03-13 (BioX, PRMU) |