Generalized Multiview Discriminative Projections With Spectral Reconstruction

In image recognition, there are a large number of small sample size problems in which the number of training samples is less than the dimension of feature vectors. For such problems, generalized multiview linear discriminant analysis (GMLDA) usually fails to achieve good learning performance for many classification tasks. With the idea of fractional order embedding, this paper proposes a new multiview feature learning method via fractional spectral modeling, namely, fractional-order generalized multiview discriminant analysis (FGMDA), which is able to subsume GMLDA as a special case. Experimental results on visual recognition have demonstrated the effectiveness of the proposed method and shown that FGMDA outperforms GMLDA.