LMDA: Local Maximum Discrimination Analysis

In this paper, we put forward a novel supervised feature extraction method based on the Linear Discrimination Analysis: Local Maximum Discrimination Analysis. Also, in order to strengthen the local learning ability of the algorithm and improve the capable of reducing dimensionality, we introduce the Local Weighted Mean to the algorithm LMDA. Better is that, there is no Small Sample Size Problem in the new proposed algorithm with the introduction of Maximum Margin Criterion. In the end, experimental results demonstrate the above advantages of the algorithm LMDA.