Heterogeneous feature space based task selection machine for unsupervised transfer learning

Publication Type:
Conference Proceeding
Proceedings - The 2015 10th International Conference on Intelligent Systems and Knowledge Engineering, ISKE 2015, 2016, pp. 46 - 51
Issue Date:
Full metadata record
© 2015 IEEE. Transfer learning techniques try to transfer knowledge from previous tasks to a new target task with either fewer training data or less training than traditional machine learning techniques. Since transfer learning cares more about relatedness between tasks and their domains, it is useful for handling massive data, which are not labeled, to overcome distribution and feature space gaps, respectively. In this paper, we propose a new task selection algorithm in an unsupervised transfer learning domain, called as Task Selection Machine (TSM). It goes with a key technical problem, i.e., feature mapping for heterogeneous feature spaces. An extended feature method is applied to feature mapping algorithm. Also, TSM training algorithm, which is main contribution for this paper, relies on feature mapping. Meanwhile, the proposed TSM finally meets the unsupervised transfer learning requirements and solves the unsupervised multi-task transfer learning issues conversely.
Please use this identifier to cite or link to this item: