generalization.m

matlab实现主成分分析PAC神经网络的程序

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%% Copy Right:  the Computational Intelligence Laboratory 
%% System name: PCA
%% File name:   Generalization.m
%% Description: This is a simulation of the generalization capability of neural
%%              network. in the simulation, the tiger image and fruit image will
%%              is used.
%%
%% Author: 
%% Date:         4/14/2004
%% 
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

clear all;
tigerImage = imread('tigertest.bmp');


[rr,cc] = size(tigerImage);
Img = tigerImage;
mc=256;

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%  the image is digited into the 16x16384 matrix.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

r = 4; c = 4 ; p = r*c ;
tigerImage = blkM2vc(Img, [r c]) ;
[p N] = size(tigerImage) ; 

Xm = mean(tigerImage')' ;
Xmax=max(max(abs(tigerImage)));
tigerImage = tigerImage - Xm(:, ones(1, N)) ;
tigerImage = tigerImage/max(max(abs(tigerImage))) ;

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%  the original tiger image is shows.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
tempXX=tigerImage*Xmax;
tempXX=tempXX+Xm(:, ones(1, N)) ;
Imr = vc2blkM(tempXX, r, rr) ;
Imr = round(mc*Imr/max(max(Imr))) ;
figure(1), set(1,'Name','the original tiger image'),image(Imr), colormap(gray(mc))



%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%  the neural network is trained by the input data of the tiger image.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

m = 2 ; % number of neurons
time(m)=rem(Now,1);
W = 0.6*(rand(m, p)-0.5); % weight initialisation
lda=rand(m,1);
er = .0001; % the length convergence error
e0=tigerImage;

for j=1:m


    for s=1:16
    [rs rn] = sort(rand(1, N)) ;
    for n=1:N
        if n==1
            eta(j)=sum(sum(e0.^2))/N;    
        end
        x = e0(:, rn(n)) ; % randomised selection of patterns
        y = W(j,:)*x ;
        eta(j)=y^2+eta(j);
        dW(j,:)=(y/eta(j))*(x' - W(j,:)*y);
        W(j,:)=W(j,:)+dW(j,:);
        if  sqrt(sum(dW(j,:).^2))<er 
            break; 
        end
    end
    if  sqrt(sum(dW(j,:).^2))<er 
            break; 
    end
    end
    y=W(j,:)*e0;
    e0=e0-W(j,:)'*y;
end

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%  reading the fruit image.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

fruitImage = imread('fruit.bmp');
[rr,cc] = size(fruitImage);

Img = fruitImage;
mc=256;

r = 4; c = 4 ; p = r*c ;
fruitImage = blkM2vc(Img, [r c]) ;
[p N] = size(fruitImage) ; 

Xm = mean(fruitImage')' ;
Xmax=max(max(abs(fruitImage)));
fruitImage = fruitImage - Xm(:, ones(1, N)) ;
fruitImage = fruitImage/max(max(abs(fruitImage))) ;
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%  showing the original fruit image.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

XX=fruitImage*Xmax;
XX=XX+Xm(:, ones(1, N)) ;
Imr = vc2blkM(XX, r, rr) ;
Imr = round(mc*Imr/max(max(Imr))) ;
figure(2),set(2,'Name','the original fruit image'), image(Imr), colormap(gray(mc))

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%  extrating the principal component of the fruit data set.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%


yy = W*fruitImage ;



%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%  Reconstructing the fruit image.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

Xr = W'*yy;

Xr=Xr*Xmax;
Xr = Xr + Xm(:, ones(1, N)) ;
Imr = vc2blkM(Xr, r, rr) ;
Imr = round(mc*Imr/max(max(Imr))) ;
Imr = round(Imr) ;
figure(3),set(3,'Name','the generalization result'), image(Imr), colormap(gray(mc))

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%% this is end.