Unsupervised metric learning with synthetic examples
Distance Metric Learning (DML) involves learning an embedding that brings similar examples closer while moving away dissimilar ones. Existing DML approaches make use of class labels to generate constraints for metric learning. In this paper, we address the less-studied problem of learning a metric i...
| Published in: | Proceedings of the AAAI Conference on Artificial Intelligence |
|---|---|
| Main Authors: | , , |
| Other Authors: | , |
| Format: | Other Non-Article Part of Journal/Newspaper |
| Language: | English |
| Published: |
Association for the Advancement of Artificial Intelligence (AAAI)
2020
|
| Subjects: | |
| Online Access: | https://research.monash.edu/en/publications/57ff31f8-f673-4448-8a1d-1a51ede2255d https://doi.org/10.1609/aaai.v34i04.5795 https://researchmgt.monash.edu/ws/files/381951922/351816113_oa.pdf http://www.scopus.com/inward/record.url?scp=85106445643&partnerID=8YFLogxK |
| Summary: | Distance Metric Learning (DML) involves learning an embedding that brings similar examples closer while moving away dissimilar ones. Existing DML approaches make use of class labels to generate constraints for metric learning. In this paper, we address the less-studied problem of learning a metric in an unsupervised manner. We do not make use of class labels, but use unlabeled data to generate adversarial, synthetic constraints for learning a metric inducing embedding. Being a measure of uncertainty, we minimize the entropy of a conditional probability to learn the metric. Our stochastic formulation scales well to large datasets, and performs competitive to existing metric learning methods. |
|---|