%======================================================================
        % Copyright (C) 2005
        % All rights reserved.

        % File name: FriedmanI.m 
        % Notes:     應用RBF網絡對 Friedman #1 進行函數逼近
        %            檢驗反向傳播的逼近效果
        %            最簡單的簡單的梯度下降算法

        % Version: 1.0
        % Author:  bug zhang
        % Date:    2005.1
        %======================================================================     
        
        close all	%關閉所有的窗口
        clear all	%清除所有的變量
        k=input('訓練集點數=k,k=[3000]   ');
        if isempty(k),
           k=3000;
        end;
        L=input('最大隱層神經元個數=L,L=[300]   ');
        if isempty(L),
           L=300;
        end;
        S=input('歸一化參數=S,S=[1]   ');
        if isempty(S),
           S=1;
        end;

        X1=rand(1,k);
        X2=rand(1,k);
        X3=rand(1,k);
        X4=rand(1,k);
        X5=rand(1,k);
        X=[X1;X2;X3;X4;X5];
        Y=10.*sin(pi.*X1.*X2)+20*(X3-0.5).^2+10.*X4+5.*X5;
        
        net=newrb(X,Y,0,S,L);
        
        X1=rand(1,3000);
        X2=rand(1,3000);
        X3=rand(1,3000);
        X4=rand(1,3000);
        X5=rand(1,3000);
        X=[X1;X2;X3;X4;X5];
        Y1=10.*sin(pi.*X1.*X2)+20.*(X3-0.5).^2+10.*X4+5.*X5;
        Y2=sim(net,X);
        
        R=linspace(min(Y1),max(Y1),1000);
        T=R;
        
        figure(2)
        plot(R,T,'-k',Y1,Y2,'+r')