NEAREST SUBSPACE PRESERVING FOR FEATURE EXTRACTION METHOD
To solve the problem of insufficient local confidence in the definition of manifold learning method,by maintaining local internal and spatial relationships to capture low dimensional manifolds of data,we propose a feature extraction method based on the nearest subspace preserving.Every sample point and its K nearest neighbors in the data were treated as a locality,and a nearest subspace was stretched.The Gram determinant was used to measure the volume of all the nearest subspaces.The volume was normalized and integrated into the model of the locality preserving projections algorithm.The experimental results of clustering and classification on real data prove that the features extracted by our approach are more discriminative.
万方数据
维普
