18 June 2018 Iterative landmark selection and subspace alignment for unsupervised domain adaptation
Author Affiliations +
Abstract
Domain adaptation (DA) solves a learning problem in a target domain by utilizing the training data in a different but related source domain, when the two domains have the same feature space and label space but different distributions. An unsupervised DA approach based on iterative landmark selection and subspace alignment (SA) is proposed. The proposed method automatically selects source landmarks from the source domain and iteratively selects target landmarks from the target domain. These well-selected landmarks accurately reflect the similarity between the two domains and are applied to kernel projection of both source and target samples onto a common subspace, where SA is performed. In each iteration, target labels are updated by a classifier that is retrained with the source samples aligned with the target domain. Thus, the distribution of the selected target landmarks gradually approximates the distribution of the source domain. During landmark selection, the quadratic optimization functions are constrained such that the proportions of selected samples per class remain the same as in the original domain, which makes the problem easy to solve and avoids setting hyperparameters. Comprehensive experimental results show that the proposed method is effective and outperforms state-of-the-art adaptation methods.
© 2018 SPIE and IS&T
Ting Xiao, Ting Xiao, Peng Liu, Peng Liu, Wei Zhao, Wei Zhao, Xianglong Tang, Xianglong Tang, } "Iterative landmark selection and subspace alignment for unsupervised domain adaptation," Journal of Electronic Imaging 27(3), 033037 (18 June 2018). https://doi.org/10.1117/1.JEI.27.3.033037 . Submission: Received: 16 March 2018; Accepted: 30 May 2018
Received: 16 March 2018; Accepted: 30 May 2018; Published: 18 June 2018
JOURNAL ARTICLE
14 PAGES


SHARE
RELATED CONTENT

Saliency detection via background features
Proceedings of SPIE (August 08 2018)
A unified framework for PCA, LDA, and LPP
Proceedings of SPIE (June 01 2012)
Gaussian mass optimization for kernel PCA parameters
Proceedings of SPIE (September 30 2011)
Very fast recognition of GIRO check forms
Proceedings of SPIE (April 13 1993)

Back to Top