Regional data set linear separable algorithm

Abstract

The invention discloses a regional data set linear separable algorithm which comprises the following steps: (1) after importing a corresponding data set, carrying out coordinate transformation; (2) projecting points in each quadrant; (3) if only one point exists, outputting a result: linear separability; and (4) if two points exist, further judgment is needed. If the two points are in the two connected quadrants, the output is linearly separable, otherwise, the output is unable to judge; and (5) if three points exist, further operation is needed. Compared with the prior art, the method has theadvantages that on the basis of coordinate axes, the separable algorithm takes two-dimensional and three-dimensional data as an example (other dimension data can be subjected to dimension reduction processing) to carry out linear separability on a comparison point and a regional data set. And the algorithm is simpler, less in time consumption and more convenient.

Description

technical field [0001] The invention relates to the fields of machine vision and neural network, in particular to a linearly separable algorithm for a region data set. Background technique [0002] For the solution of the nonlinear classification model problem, although the existence of the problem solution has been proved theoretically, it is still difficult to find the solution of the problem in practical applications. These difficulties are mainly reflected in two aspects: 1) The choice of kernel function in SVM model is uncertain; 2) In neural network model, the design of the number of hidden layer neurons is uncertain. All in all, these difficulties still arise from whether the mapping from the classification task to the hidden layer space is linearly separable. Contents of the invention [0003] The technical problem to be solved by the present invention is to provide a linearly separable algorithm for a region data set aiming at the above problems. [0004] In ord...

AI Technical Summary

Problems solved by technology

[0002] For the solution of the nonlinear classification model problem, although the existence of the problem solution has been proved theoretically, it is still difficult to find the solution of the problem in practical applications.

These difficulties are mainly reflected in two aspects: 1) The choice of kernel function in the SVM model is uncertain; 2) In the neural network model, the design of the number of hidden layer neurons is uncertain

All in all, these difficulties still arise from whether the mapping from the classification task to the hidden layer space is linearly separable

Images