| abstract |
Neighbour is supervised the invention discloses a kind of having based on kernel function and keeps embedding grammar, and the present invention is compared with differentiating neighbour's embedded mobile GIS, the present invention energy unbalanced problem of processing data sample distribution, and discrimination is higherï¼›It will be trained by kernel function and test sample transform to non-linear space, then using the classification information of known training examples, learning training obtains dimensionality reduction eigenmatrix, so that sample has more preferable separability in subspace is differentiated. |