ex1d1.m

新的神经网络算法源程序

%==============================================================
%    Ex1d1: An example for using WaveNet,  1-D Interpolation
%==============================================================

% By Qinghua Zhang. March, 1994.

help ex1d1

echo on

% The training data
%===================

% Input patterns
x = [-0.9588   -0.8611   -0.7614   -0.6790   -0.6074   -0.4920   -0.3364 ...
     -0.2987   -0.1644   -0.1520    0.0473    0.1727    0.1766    0.2596 ...
      0.3616    0.5276    0.5697    0.6961    0.7437    0.9412    0.9696]';

% Output patterns
y = [-0.9602   -0.5770   -0.0729    0.3771    0.6405    0.6600    0.4609 ...
      0.1336   -0.2013   -0.4344   -0.5000   -0.3930   -0.1647    0.0988 ...
      0.3072    0.3960    0.3449    0.1816   -0.0312   -0.2189   -0.3201]';

% The data are depicted in the graphics window.

clf; plot(x,y,'o'); 
xlabel('x'); ylabel('y'); title('The data to be interpolated');

% We are going to use a WaveNet to interpolate the data.
% You may set the number of wavelets in the network to 3
% or choose it on-line with the aid of the FPE criterion.

% Call the training procedure
%=============================

Key = keymenu('Make your choice','Use 3 wavelets','On-line choice');

if Key==1
  th = wnetreg(y, x, 3, 20, 2); % Use 3 wavelets, 20 iterations, 
                                         % Initialization mode 2
else
  th = wnetreg(y, x, [], 20, 2); % Number of wavelets not determined.
end

% Examine the result of the interpolation
%=========================================

% Evaluate the network on a finer x-sequence

xg = (-1:.01:1)';
yg = wavenet(xg, th);

plot(x,y,'o', xg,yg); xlabel('x'); ylabel('y');

% The result of the interpolation is ploted in the graphics window.

echo off