eduzhai > Applied Sciences > Engineering >

Dimensionality Reduction via Diffusion Map Improved with Supervised Linear Projection

  • king
  • (0) Download
  • 20210506
  • Save

... pages left unread,continue reading

Document pages: 6 pages

Abstract: When performing classification tasks, raw high dimensional features oftencontain redundant information, and lead to increased computational complexityand overfitting. In this paper, we assume the data samples lie on a singleunderlying smooth manifold, and define intra-class and inter-class similaritiesusing pairwise local kernel distances. We aim to find a linear projection tomaximize the intra-class similarities and minimize the inter-class similaritiessimultaneously, so that the projected low dimensional data has optimizedpairwise distances based on the label information, which is more suitable for aDiffusion Map to do further dimensionality reduction. Numerical experiments onseveral benchmark datasets show that our proposed approaches are able toextract low dimensional discriminate features that could help us achieve higherclassification accuracy.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×