A novel method called orthogonal marginal Fisher analysis (OMFA) for dimensionality reduction is proposed.It is proved that,OMFA can maximize the intra-class compactness while minimizing the inter-class separability.The algorithm is used to face recognition,and the numerical experimental results show that it outperforms the original MFA.%提出了一种新的名为正交边际Fisher分析的方法用于降维,并且证明了了该方法可以同时最大化类内数据的紧致性和最小化类外数据的分离度.在人脸识别的实验中,表明该方法的性能比原来的边际Fisher分析有较大的提高.
展开▼
