Based on the steepest descent method and the dicho
2016-08-23
0 0 0
no vote
Other
Earn points
Gradient method of nonlinear optimal value can quickly converge to near optimal solution, since there are rules to participate in training weighting parameters, the reference value, the result of confidence, so when demand fell in the direction of the gradient method
Using the definition of partial derivatives to solve the constant decrease in the time step size is calculated using the two points on the step for processing, so that the objective function Mse has reached the final of the accuracy requirements
Comparison: machine configuration 32bit / win7 Mse = 0.00018
Parameters in matlab training function Fmincon training time is 15 minutes
Based in matlab steepest descent method and the dichotomy of BRB parameter training time is 13 seconds
Using C-based steepest descent method and the dichotomy of BRB parameters training time just 0.8 seconds
(Author: wwk)
Using the definition of partial derivatives to solve the constant decrease in the time step size is calculated using the two points on the step for processing, so that the objective function Mse has reached the final of the accuracy requirements
Comparison: machine configuration 32bit / win7 Mse = 0.00018
Parameters in matlab training function Fmincon training time is 15 minutes
Based in matlab steepest descent method and the dichotomy of BRB parameter training time is 13 seconds
Using C-based steepest descent method and the dichotomy of BRB parameters training time just 0.8 seconds
(Author: wwk)
c
Related Source Codes
PClite
0
0
no vote
generic embeded FTP Client
0
0
no vote
UDP Test Demo
0
0
no vote
TMS320F28335 DSP transmits data to EEPROM through
0
0
no vote
TMS320F28335 CAN communication source code
0
0
no vote
No comment