DC Field | Value | Language |
---|---|---|
dc.contributor.author | Li, Zhun | ko |
dc.contributor.author | Ko, ByungSoo | ko |
dc.contributor.author | Choi, Ho-Jin | ko |
dc.date.accessioned | 2019-10-01T00:20:04Z | - |
dc.date.available | 2019-10-01T00:20:04Z | - |
dc.date.created | 2019-07-01 | - |
dc.date.issued | 2019-09 | - |
dc.identifier.citation | PEER-TO-PEER NETWORKING AND APPLICATIONS, v.12, no.5, pp.1358 - 1368 | - |
dc.identifier.issn | 1936-6442 | - |
dc.identifier.uri | http://hdl.handle.net/10203/267705 | - |
dc.description.abstract | To facilitate the utilization of large-scale unlabeled data, we propose a simple and effective method for semi-supervised deep learning that improves upon the performance of the deep learning model. First, we train a classifier and use its outputs on unlabeled data as pseudo-labels. Then, we pre-train the deep learning model with the pseudo-labeled data and fine-tune it with the labeled data. The repetition of pseudo-labeling, pre-training, and fine-tuning is called naive semi-supervised deep learning. We apply this method to the MNIST, CIFAR-10, and IMDB data sets, which are each divided into a small labeled data set and a large unlabeled data set by us. Our method achieves significant performance improvements compared to the deep learning model without pre-training. We further analyze the factors that affect our method to provide a better understanding of how to utilize naive semi-supervised deep learning in practical application. | - |
dc.language | English | - |
dc.publisher | SPRINGER | - |
dc.title | Naive semi-supervised deep learning using pseudo-label | - |
dc.type | Article | - |
dc.identifier.wosid | 000484477700024 | - |
dc.identifier.scopusid | 2-s2.0-85058144754 | - |
dc.type.rims | ART | - |
dc.citation.volume | 12 | - |
dc.citation.issue | 5 | - |
dc.citation.beginningpage | 1358 | - |
dc.citation.endingpage | 1368 | - |
dc.citation.publicationname | PEER-TO-PEER NETWORKING AND APPLICATIONS | - |
dc.identifier.doi | 10.1007/s12083-018-0702-9 | - |
dc.contributor.localauthor | Choi, Ho-Jin | - |
dc.contributor.nonIdAuthor | Li, Zhun | - |
dc.contributor.nonIdAuthor | Ko, ByungSoo | - |
dc.description.isOpenAccess | N | - |
dc.type.journalArticle | Article | - |
dc.subject.keywordAuthor | Deep learning | - |
dc.subject.keywordAuthor | Semi-supervised learning | - |
dc.subject.keywordAuthor | Pseudo-label | - |
dc.subject.keywordAuthor | Pre-training | - |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.