Unlock instant, AI-driven research and patent intelligence for your innovation.

Anti-noise gradient direction neurokinetics algorithm based on parallel technology

A technology of gradient direction and neurodynamics, applied in the direction of neural learning methods, neural architecture, biological neural network models, etc., can solve the problems of increasing calculation time consumption, improve calculation efficiency, avoid calculation time consumption, and strong noise tolerance and the effect of computational accuracy

Pending Publication Date: 2021-06-11
GUANGDONG OCEAN UNIVERSITY
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0006] For above-mentioned existing problems, the present invention aims to provide a kind of anti-noise gradient direction neural dynamics algorithm based on parallel technology, by utilizing the anti-noise gradient direction neural network method based on parallel technology to solve Sylvester equation, and traditional gradient solving method Compared with this method, when solving the Sylvester equation with noise, it can avoid the problem of increasing the time-consuming calculation of the pseudo-inverse, greatly improve the calculation efficiency, and enhance the anti-noise ability of the algorithm by introducing the integral term, which has good robustness, High computing efficiency and good noise immunity

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Anti-noise gradient direction neurokinetics algorithm based on parallel technology
  • Anti-noise gradient direction neurokinetics algorithm based on parallel technology
  • Anti-noise gradient direction neurokinetics algorithm based on parallel technology

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0091] Embodiment 1: S301. When Sylverster equation (1) has the situation of analytical solution, the coefficient matrix that the present invention provides equation is respectively: P=[sin(3) cos(3);-cos(3) sin(3) ], Q=0 and L=-I; then, the analytical solution X of the equation is obtained by simple algebraic operations:

[0092]

[0093] Moreover, the corresponding residual ||E(t)|| F Defined as E(t)=PX(t)-X(t)Q+L, its corresponding vectorized residual is expressed as e(t), and this variable is used as the evaluation index of convergence; among them, the following mark F Represents the Frobenius norm of the matrix;

[0094] The computer simulation results are as figure 2 and image 3 ;Depend on figure 2 It can be seen that from the randomly generated initial state X 0 ∈ [-1, 1] 2×2 At the beginning, in the case of adding different types of noise (constant noise, linear noise and random noise), the anti-noise gradient direction neural dynamics (NTGON) algorithm pr...

Embodiment 2

[0097] Embodiment 2: Next, the present invention provides the example that a Sylvester equation does not have analytic solution, compares two kinds of algorithms aspect performance in numerical solution solving equation, detailed analysis is under the convergence performance under various noise conditions:

[0098] S302. When the Sylverster equation does not have an analytical solution, the coefficient matrix of the equation provided by the present invention is respectively: P=[sin(4) cos(4);-cos(4) sin(4)], Q=[2 0 ;0 3] and L=[sin(1) cos(1);-cos(1) sin(1)];

[0099] Figure 4-6 Presented the NTGON algorithm proposed by the present invention and the calculation performance comparison situation of the original GNN algorithm for solving the Sylvester equation; at the same time, in order to understand the influence of the convergence scaling factor of the algorithm on the calculation performance, the present invention combines three kinds of noise (constant noise ( Figure 4 ), ...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention discloses an anti-noise gradient direction neurokinetics algorithm based on a parallel technology, and the algorithm comprises the steps: S1, proposing a Sylvester equation, and obtaining a linear equation set used for solving the Sylvester equation; S2, introducing a scalar error function and a classical linear gradient neural network (GNN) algorithm according to the linear equation set, and obtaining an anti-noise gradient direction neurokinetics algorithm used for solving the Sylvester equation. According to the algorithm, an anti-noise gradient direction neural network method based on the parallel technology is used for solving the Sylvester equation, compared with a traditional gradient solving method, when the Sylvester equation containing noise is solved, the problem that the calculation time consumption is increased due to pseudo-inverse solving can be avoided, the calculation efficiency is greatly improved, and the anti-noise capacity of the algorithm is enhanced by introducing an integral term; the algorithm has the characteristics of good robustness, high calculation efficiency and good anti-noise capability.

Description

technical field [0001] The invention relates to the technical field of matrix equation solving algorithms, in particular to an anti-noise gradient direction neural dynamics algorithm based on parallel technology. Background technique [0002] Matrix equations are widely used in many engineering and scientific fields, and play a very important role in computational mathematics and control theory; in practical problems, how to quickly and effectively deal with large linear matrix equations is often proposed and has become a current hot topic and key problem; [0003] The direct method and the iterative method are two methods for solving linear algebraic systems. The direct method is often used when solving linear systems. The advantage of the direct method is that it can obtain an accurate solution under the condition that the rounding error can be ignored; however, when the coefficient matrix When the number of conditions is large, the accuracy of the solution obtained by th...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Applications(China)
IPC IPC(8): G06F30/27G06N3/04G06N3/08G06F119/10
CPCG06N3/04G06N3/08G06F30/27G06F2119/10
Inventor 付东洋刘贝金龙肖秀春余果刘大召高兵
Owner GUANGDONG OCEAN UNIVERSITY