A method often used in pattern recognition is called Flasher linear discrimination.

The basic idea of ??Fisher's discrimination is projection. For a certain point x=(x1, x2, x3,..., xp) in P-dimensional space, find a linear function y(x) that can reduce it to a one-dimensional value: y(x

)= ∑Cjxj Then apply this linear function to transform the population of known categories in the P-dimensional space and the samples of unknown category affiliation into one-dimensional data, and then determine the ownership of the sample points of unknown affiliation based on the degree of closeness between them.

This linear function should be able to minimize the differences between sample points in the same category and maximize the differences between sample points in different categories after converting all points in the P-dimensional space into one-dimensional values.

The difference between them can achieve higher discrimination efficiency.

The idea of ??one-variance analysis is borrowed here, that is, the judgment is made based on the principle of maximizing the ratio of the mean square error between groups to the mean square error within the group.