Combining clustering and active learning for the detection and learning of new image classes
| Autor: | Luiz Fernando Sommaggio Coletta, Ayan Acharya, Eduardo R. Hruschka, Moacir Antonelli Ponti, Joydeep Ghosh |
|---|---|
| Přispěvatelé: | Universidade Estadual Paulista (Unesp), Universidade de São Paulo (USP), Univ Texas Austin |
| Rok vydání: | 2019 |
| Předmět: |
0209 industrial biotechnology
Active learning Image classification Computer science Cognitive Neuroscience 02 engineering and technology Machine learning computer.software_genre Clustering 020901 industrial engineering & automation Discriminative model Artificial Intelligence 0202 electrical engineering electronic engineering information engineering RECONHECIMENTO DE IMAGEM Entropy (information theory) Entropy (energy dispersal) Cluster analysis business.industry Entropy (statistical thermodynamics) Deep learning Open set Computer Science Applications Support vector machine 020201 artificial intelligence & image processing Artificial intelligence business computer |
| Zdroj: | Web of Science Repositório Institucional da UNESP Universidade Estadual Paulista (UNESP) instacron:UNESP Repositório Institucional da USP (Biblioteca Digital da Produção Intelectual) Universidade de São Paulo (USP) instacron:USP |
| ISSN: | 0925-2312 |
| DOI: | 10.1016/j.neucom.2019.04.070 |
| Popis: | Made available in DSpace on 2019-10-04T12:38:21Z (GMT). No. of bitstreams: 0 Previous issue date: 2019-09-17 Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP) Discriminative classification models often assume all classes are available at the training phase. As such models do not have a strategy to learn new concepts from available unlabeled instances, they usually work poorly when unknown classes emerge from future data to be classified. To address the appearance of new classes, some authors have developed approaches to transfer knowledge from known to unknown classes. Our study provides a more flexible approach to learn new (visual) classes that emerge over time. The key idea is materialized by an iterative classifier that combines Support Vector Machines with clustering via an optimization algorithm. An entropy and density-based selection strategy explores label uncertainty and high-density regions from unlabeled data to be classified. Selected instances from new classes are submitted to get labels and then used to improve the model. The proposed image classifier is consistently better than approaches that select instances randomly or from clusters. We also show that features obtained via Deep Learning methods improve results when compared with shallow features, but only using our selection strategy. Our approach requires fewer iterations to learn new classes, thereby significantly saving labeling costs. (C) 2019 Elsevier B.V. All rights reserved. Sao Paulo State Univ, Sch Sci & Engn, Tupa, SP, Brazil Univ Sao Paulo, Inst Math & Comp Sci, Sao Carlos, SP, Brazil Univ Sao Paulo, Dept Comp Engn & Digital Syst, Sao Carlos, SP, Brazil Univ Texas Austin, Dept Elect & Comp Engn, IDEAL, Austin, TX 78712 USA Univ Texas Austin, Dept Elect & Comp Engn, Machine Learning Res Grp, Austin, TX 78712 USA Univ Texas Austin, Dept Elect & Comp Engn, Austin, TX 78712 USA Sao Paulo State Univ, Sch Sci & Engn, Tupa, SP, Brazil FAPESP: 2017/00357-7 |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
Full Text from ScienceDirect