Abstract:

Lie group theory is the fundamental representation theory of the transformation space. Presently, there are only a few classifiers for Lie group matrix data and they do not work well for multiclass classification tasks. Taking handwritten digits as application background, the paper introduces the support vector machine (SVM) to handle Lie group matrix data. Since the Lie group data has their matrix form, we design a matrix Gaussian kernel function, making the SVM to handle Lie group matrix data. Experimental results show that the SVM performs well on the Lie group matrix data.

Key words: support vector machine;Lie group;matrix data