Transfer Domain Class Clustering for Unsupervised Domain Adaptation
In this paper, we propose a transfer domain class clustering (TDCC) algorithm to address the unsupervised domain adaptation problem, in which the training data (source domain) and the test data (target domain) follow different distributions. TDCC aims to derive new feature representations for source and target in a latent subspace to simultaneously reduce the distribution distance between two domains, which helps transfer the source knowledge to the target domain effectively, and enhance the class discriminativeness of data as much as possible by minimizing the intra-class variations, which can benefit the final classification a lot. The effectiveness of TDCC is verified by comprehensive experiments on several cross-domain datasets, and the results demonstrate that TDCC is superior to the competitive algorithms.
KeywordsFeature learning Distribution adaptation Domain adaptation Transfer learning
This research is supported by the CRRC Major Scientific Projects under Grant No. 2106CKZ206-1 and National Key R&D Program under Grant No. 2016YFB1200203.
- 2.Long M, Wang J, Ding G et al (2014) Transfer joint matching for unsupervised domain adaptation. In: Proceedings of the IEEE conference on computer vision and pattern recognition, pp 1410–1417Google Scholar
- 3.Long M, Wang J, Ding G et al (2013) Transfer feature learning with joint distribution adaptation. In: Proceedings of the IEEE international conference on computer vision, pp 2200–2207Google Scholar
- 4.Li S, Song S, Huang G (2016) Prediction reweighting for domain adaptation. IEEE Trans Neural Netw Learn SystGoogle Scholar
- 8.Gretton A, Borgwardt KM, Rasch MJ et al (2013) A kernel two-sample test. J Mach Learn Res 13:723–773Google Scholar
- 9.Jolliffe I (2002) Principal component analysis. WileyGoogle Scholar
- 11.Gong B, Shi Y, Sha F et al (2012) Geodesic flow kernel for unsupervised domain adaptation. In: Computer vision and pattern recognition (CVPR), 2012 IEEE Conference on IEEE, pp 2066–2073Google Scholar
- 12.Wang H, Wang W, Zhang C et al (2014) Cross-domain metric learning based on information theory. AAAI, pp 2099–2105Google Scholar