Large-margin multi-view Gaussian process for image classification

Publisher:
ACM
Publication Type:
Conference Proceeding
Citation:
Proceedings of the Fifth International Conference on Internet Multimedia Computing and Service, 2013, pp. 7 - 12
Issue Date:
2013-01
Full metadata record
Files in This Item:
Filename Description Size
Thumbnail2013003402OK.pdf2.73 MB
Adobe PDF
In image classification, the goal is to decide whether an image belongs to a certain category or not. Multiple features are usually employed to comprehend the contents of images substantially for the improvement of classification accuracy. However, it also brings in some new problems that how to effectively combine multiple features together, and how to handle the high-dimensional features from multiple views given the small training set. In this paper, we present a large-margin Gaussian process approach to discover the latent space shared by multiple features. Therefore, multiple features can complement each other in this low-dimensional latent space, which derives a strong discriminative ability from the large-margin principle, and then the following classification task can be effectively accomplished. The resulted objective function can be efficiently solved using the gradient descent techniques. Finally, we demonstrate the advantages of the proposed algorithm on real-world image datasets for discovering discriminative latent space and improving the classification performance.
Please use this identifier to cite or link to this item: