Non-parametric subspace classifier, such as collaborative representation based classification, sparse representation based classification obtains superior performance to conventional parametric model method, such as support vector machine and softmax regression, for visual classification. Recently, a probabilistic collaborative representation based classifier, which utilizes a hybrid representation (shared representation and class specific representation) to a test sample, leading to state-of-the-art classification performance. However, the probabilistic collaborative representation based classification does not consider the nonlinear characteristics hidden in visual features. In the paper, we propose to utilize kernel technique to extend the probabilistic collaborative presentation based classification method. Experimental results on several benchmark datasets demonstrate that our propose method obtains favourable classification performance.
CITATION STYLE
Meng, J., Wang, Y., & Liu, B. D. (2018). Probabilistic collaborative representation with kernels for visual classification. In Communications in Computer and Information Science (Vol. 819, pp. 392–402). Springer Verlag. https://doi.org/10.1007/978-981-10-8530-7_38
Mendeley helps you to discover research relevant for your work.