When the dimension of data is larger than the number of samples, linear discriminant analysis (LDA) cannot be applied directly to high-dimensional data. This case is called small-sample-size (SSS) or under sampled problem.
To solve this problem, some local learning based various image clustering models are proposed.
Pak Kwang Jun, a lecturer at the Faculty of Applied Mathematics, has proposed a local learning based exponential regularized discriminant clustering model.
In the proposed local exponential regularized discriminant clustering (LERDC) model, local scatter matrices of regularized discriminant model are projected in the exponential domain in order to handle the SSS problem of LDA. In the proposed LERDC model, for each image, the local image matrix is constructed comprising k nearest neighbor images, and the local exponential regularized discriminant model (LERDM) is devised to evaluate the clustering results for the images in the local image matrix.
To verify his method, he compared it with existing state-of-the-art local learning based clustering approaches. The results showed that the proposed LERDC model achieved a comparable clustering performance to that of the near competitor LDMGI model which is based on LDA.
© 2021 KumChaek University of Technology