Maximum Scatter Difference,Large Margin Linear Projection and Support Vector Machines
Song Feng, Cheng Ke · Acta Automatica Sinica · 2004
A modified Fisher discriminant is proposed at first.Then maximum scatter dif-ference classifier(MSDs)which is based on the new discriminant is derived.It is showed thatwhen parameter C in the MSDs is approaching infinity,a new kind of classifier called largemargin linear projection classifier(LMLP)can be obtained.Theoretical analysis indicatesthat LMLP is a special case of linear support vector machines when the pattern samples arelinearly separable.Experimental results conducted on the ORL and NUST603 datasets showthat the MSDs and LMLP are better than traditional linear discriminant analysis methodssuch as Foley-Sammon linear discriminant analysis,and can compete with linear supportvector machines and uncorrelated linear discriminant analysis.