nndadapt.m

来自「神经网络设计书籍配套源码」· M 代码 · 共 34 行

function wprime = nndadapt(t,w)
%NNDADAPT Calculates the derivative for adaptive weights of the Grossberg network
%
%	NNDADAPT(t,w2)
%	  t - Current time
%	  w - Weights (w(1)=w1,1, w(2)=w12, w(3)=w21, w(4)=w22)
%   Returns dw.

% First Version, 8-31-95.

%==================================================================

global n1_1;
global n1_2;
global lr;

if (rem(fix(t/0.2),2)==0),
  n1 = n1_1;
  n2 = [1; 0];
else
  n1 = n1_2;
  n2 = [0; 1];
end

w = reshape(w,2,2);

if lr == 1
  wprime = (4*n2*ones(1,2)).*(ones(2,1)*n1'-w);
else
  wprime = 4*n2*n1' - 2*w;
end

wprime = reshape(wprime,4,1);