English

Joint cross-domain classification and subspace learning for unsupervised adaptation

Computer Vision and Pattern Recognition 2015-04-30 v3 Machine Learning

Abstract

Domain adaptation aims at adapting the knowledge acquired on a source domain to a new different but related target domain. Several approaches have beenproposed for classification tasks in the unsupervised scenario, where no labeled target data are available. Most of the attention has been dedicated to searching a new domain-invariant representation, leaving the definition of the prediction function to a second stage. Here we propose to learn both jointly. Specifically we learn the source subspace that best matches the target subspace while at the same time minimizing a regularized misclassification loss. We provide an alternating optimization technique based on stochastic sub-gradient descent to solve the learning problem and we demonstrate its performance on several domain adaptation tasks.

Keywords

Cite

@article{arxiv.1411.4491,
  title  = {Joint cross-domain classification and subspace learning for unsupervised adaptation},
  author = {Basura Fernando and Tatiana Tommasi and Tinne Tuytelaars},
  journal= {arXiv preprint arXiv:1411.4491},
  year   = {2015}
}

Comments

Paper is under consideration at Pattern Recognition Letters

R2 v1 2026-06-22T07:01:30.571Z