Method for calculating feature value with largest sum of rear part and imaginary part of real matrix based on neural network

Abstract

The invention discloses a method for calculating a feature value with a largest sum of a rear part and an imaginary part of a real matrix based on a neural network. The method comprises: a complex neural network model for calculating a feature value with a largest sum of a rear part and an imaginary part of a non-zero real matrix and a feature vector is constructed; a preset iterative condition, an initial model parameter and an initial feature vector of the non-zero real matrix are obtained; according to the preset iterative condition, the initial model parameter and the initial feature vector, a feature vector xi of the non-zero real matrix is obtained by iterative calculation of the complex neural network model; when the feature vector xi is equal to zero, a model parameter u s adjustedaccording to a set increment and iterative calculation of the complex neural network model is carried out until the feature vector xi is not equal to zero; and when (lambda<R>m+lambdam)+u is larger than or equal to 0, a feature vector corresponding to the feature value of the largest sum of the rear part and the imaginary part is obtained; and according to the outputted feature vector xi, a feature value with the largest sum of the rear part and the imaginary part among feature values in the non-zero real matrix A is calculated.

Description

technical field [0001] The invention relates to the technical field of information processing, in particular to a method for calculating the real and imaginary parts of a real matrix and the largest eigenvalue based on a neural network. Background technique [0002] In the process of information processing and data retrieval in the prior art, in order to obtain the most accurate information data or retrieval data, most of the input information will be converted into a matrix during the processing process, and by obtaining the modulus of the matrix or all the eigenvalues ​​of the matrix The real part imaginary part and the largest eigenvalue can quickly find the information features that need to be accurately obtained. [0003] In the prior art, related research on obtaining matrix eigenvalues ​​and eigenvectors first appeared in the 1980s. OjaE proposed to use neural networks to calculate eigenvalues ​​and eigenvectors of symmetric matrices, which is mainly for the eigenvalu...

AI Technical Summary

Problems solved by technology

However, recent papers have not found a neural network algorithm that simultaneously focuses on the sum of the imaginary and real parts of a general real matrix

Images