A Model of Data Reduction Based on Tensor Field

There has been much interest in the use of mutilinear algebra as a general technique of dimension reduction of large amounts of Tensor data in recent years. Some tensor based methods has been successfully applied to image processing signal processing and Web search. The natural forms described by tensor are kept in arbitrary coordinates and according to this invariability of tensor it is able to deal with data under the framework of tensor field. In this paper a tensor field learning model combining available algorithms of dimension reduction based on tensor is presented. The model reveals that some tensor features could be acquired from the original data and the representation of reduction data is obtained on the basis of tensor field transformation implemented by an algorithm. An example refer to reduction of image data is given and the result shows the remarkable effect of the transformation algorithm.