Similarity-Based Framework for Unsupervised Domain Adaptation: Peer Reviewing Policy for Pseudo-Labeling
The inherent dependency of deep learning models on labeled data is a well-known problem and one of the barriers that slows down the integration of such methods into different fields of applied sciences and engineering, in which experimental and numerical methods can easily generate a colossal amount...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2023-10-01
|
| Series: | Machine Learning and Knowledge Extraction |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2504-4990/5/4/74 |
| _version_ | 1797380238680784896 |
|---|---|
| author | Joel Arweiler Cihan Ates Jesus Cerquides Rainer Koch Hans-Jörg Bauer |
| author_facet | Joel Arweiler Cihan Ates Jesus Cerquides Rainer Koch Hans-Jörg Bauer |
| author_sort | Joel Arweiler |
| collection | DOAJ |
| description | The inherent dependency of deep learning models on labeled data is a well-known problem and one of the barriers that slows down the integration of such methods into different fields of applied sciences and engineering, in which experimental and numerical methods can easily generate a colossal amount of unlabeled data. This paper proposes an unsupervised domain adaptation methodology that mimics the peer review process to label new observations in a different domain from the training set. The approach evaluates the validity of a hypothesis using domain knowledge acquired from the training set through a similarity analysis, exploring the projected feature space to examine the class centroid shifts. The methodology is tested on a binary classification problem, where synthetic images of cubes and cylinders in different orientations are generated. The methodology improves the accuracy of the object classifier from 60% to around 90% in the case of a domain shift in physical feature space without human labeling. |
| first_indexed | 2024-03-08T20:34:29Z |
| format | Article |
| id | doaj.art-0937dc0214cf456eb40aa9f183f494b9 |
| institution | Directory Open Access Journal |
| issn | 2504-4990 |
| language | English |
| last_indexed | 2024-03-08T20:34:29Z |
| publishDate | 2023-10-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Machine Learning and Knowledge Extraction |
| spelling | doaj.art-0937dc0214cf456eb40aa9f183f494b92023-12-22T14:22:10ZengMDPI AGMachine Learning and Knowledge Extraction2504-49902023-10-01541474149210.3390/make5040074Similarity-Based Framework for Unsupervised Domain Adaptation: Peer Reviewing Policy for Pseudo-LabelingJoel Arweiler0Cihan Ates1Jesus Cerquides2Rainer Koch3Hans-Jörg Bauer4Institute of Thermal Turbomachinery, Karlsruhe Institute of Technology (KIT), 76137 Karlsruhe, GermanyInstitute of Thermal Turbomachinery, Karlsruhe Institute of Technology (KIT), 76137 Karlsruhe, GermanyArtificial Intelligence Research Institute (IIIA), CSIC, 08193 Bellaterra, SpainInstitute of Thermal Turbomachinery, Karlsruhe Institute of Technology (KIT), 76137 Karlsruhe, GermanyInstitute of Thermal Turbomachinery, Karlsruhe Institute of Technology (KIT), 76137 Karlsruhe, GermanyThe inherent dependency of deep learning models on labeled data is a well-known problem and one of the barriers that slows down the integration of such methods into different fields of applied sciences and engineering, in which experimental and numerical methods can easily generate a colossal amount of unlabeled data. This paper proposes an unsupervised domain adaptation methodology that mimics the peer review process to label new observations in a different domain from the training set. The approach evaluates the validity of a hypothesis using domain knowledge acquired from the training set through a similarity analysis, exploring the projected feature space to examine the class centroid shifts. The methodology is tested on a binary classification problem, where synthetic images of cubes and cylinders in different orientations are generated. The methodology improves the accuracy of the object classifier from 60% to around 90% in the case of a domain shift in physical feature space without human labeling.https://www.mdpi.com/2504-4990/5/4/74unsupervised domain adaptationpseudo-labelingtransfer learning |
| spellingShingle | Joel Arweiler Cihan Ates Jesus Cerquides Rainer Koch Hans-Jörg Bauer Similarity-Based Framework for Unsupervised Domain Adaptation: Peer Reviewing Policy for Pseudo-Labeling Machine Learning and Knowledge Extraction unsupervised domain adaptation pseudo-labeling transfer learning |
| title | Similarity-Based Framework for Unsupervised Domain Adaptation: Peer Reviewing Policy for Pseudo-Labeling |
| title_full | Similarity-Based Framework for Unsupervised Domain Adaptation: Peer Reviewing Policy for Pseudo-Labeling |
| title_fullStr | Similarity-Based Framework for Unsupervised Domain Adaptation: Peer Reviewing Policy for Pseudo-Labeling |
| title_full_unstemmed | Similarity-Based Framework for Unsupervised Domain Adaptation: Peer Reviewing Policy for Pseudo-Labeling |
| title_short | Similarity-Based Framework for Unsupervised Domain Adaptation: Peer Reviewing Policy for Pseudo-Labeling |
| title_sort | similarity based framework for unsupervised domain adaptation peer reviewing policy for pseudo labeling |
| topic | unsupervised domain adaptation pseudo-labeling transfer learning |
| url | https://www.mdpi.com/2504-4990/5/4/74 |
| work_keys_str_mv | AT joelarweiler similaritybasedframeworkforunsuperviseddomainadaptationpeerreviewingpolicyforpseudolabeling AT cihanates similaritybasedframeworkforunsuperviseddomainadaptationpeerreviewingpolicyforpseudolabeling AT jesuscerquides similaritybasedframeworkforunsuperviseddomainadaptationpeerreviewingpolicyforpseudolabeling AT rainerkoch similaritybasedframeworkforunsuperviseddomainadaptationpeerreviewingpolicyforpseudolabeling AT hansjorgbauer similaritybasedframeworkforunsuperviseddomainadaptationpeerreviewingpolicyforpseudolabeling |