人工智能/神经网络

开发平台：
Matlab

SOM_02.m：源码内容
							%------  SOM test-02 --------
clear
clc
echo off
% --------------- initial the parameter
out_num = 3;                  % 输出节点数目
input_num = 8;                % 输入节点数目
Epochs = 1;                  % 训练周期
positive_num = 55;       %  正常图像数目
negative_num = 75;       %  异常图像数目
sum_num = positive_num + negative_num; % 总的图像数目
% --------------- read data from the file
base_path = 'E:SOMBPdata_sourcefeature02';                      % minmax([I1';I2';I3';I4';I5';I6';I7';I8'])
I1 = load([base_path,'apen.txt']);           % 1 >> 2        1.2563    1.8150        
I2 = load([base_path,'kc.txt']);             % 0 >> 1        0.6299    0.9611
I3 = load([base_path,'mir.txt']);            % 0 >> 5        0.4802    4.0544
I4 = load([base_path,'asm.txt']);            % 0 >> 0.01    0.0004    0.0037
I5 = load([base_path,'idm.txt']);            % 0 >> 0.5       0.0663    0.2906
I6 = load([base_path,'cont.txt']);           % 0 >> 500     12.2600  337.5440
I7 = load([base_path,'ent.txt']);            % 5 >> 9        5.8675    7.9858
I8 = load([base_path,'gln.txt']);            % 10 >> 60      13.4204   53.2048
% normalize 
P = [(I1'-ones(1,sum_num));I2';(I3'./5);(I4'./0.01);I5'./0.5;(I6'./500);(I7'-ones(1,sum_num).*5)./4;(I8'-ones(1,sum_num).*10)./50];
MinMaxValue = [zeros(input_num,1),ones(input_num,1)]; 
% NEWSOM ---- create the som net 
net = newsom( MinMaxValue,[out_num]);
net.trainParam.show = 1000;
% net.parameter
% TRAIN ----- train the net
Epochs = 0;
savePath1 = 'E:SOMBPsave_result8feature02';
file = '.mat';
M = 10;%16;
train_num = [1,10,100,500,1000,2000,3000,5000,10000,20000,30000,50000,100000,200000,500000,1000000];
time_record = zeros(1,M);
mix_record = zeros(M,out_num,2);
% FN_record = zeros(1,M);
% FP_record = zeros(1,M);
for(i=1:M)
% train 
    if (i==1)
        Epochs = 1;
    else
        Epochs = train_num(i) - train_num(i-1);
    end
    net.trainParam.epochs = Epochs;
    %  train
    begin_time = clock;
    [net,tr,Y,E,Pf,Af] = train(net,P);
    cost_time = etime(clock,begin_time)
    time_record(i) = cost_time;
    %  my sim   
    SOMresult = dist(net.IW{1},P);
    [minValue minNum] = min(SOMresult);
    % give some test (stastical test)
    mixture = zeros(out_num,2);                         %  the positve and negative in classes
    for(k=1:out_num)
        temp = 0;
        for(j=1:positive_num)
            if( minNum(j)==k)
                temp = temp + 1;
            end
        end
        mixture(k,1) = temp;
        temp = 0;
        for(j=(positive_num+1):sum_num)
            if( minNum(j)==k)
                temp = temp + 1;
            end
        end
        mixture(k,2) = temp;
    end
    mix_record(i,:,:) = mixture(:,:); 
       
% OUTPUT ----- out put the som data which made as the bp input data
    savePath2 = [savePath1 'SOMresult' num2str(train_num(i)) file];
    save(savePath2, 'SOMresult','mixture', 'net', 'tr', 'Y' ,'E', 'Pf','Af','P'); 
end
    savePath3 = [savePath1 'statis_som'  file];
    save(savePath3,'time_record','mix_record');
  
    
 % another
 SOM_03.m;
    
%     save(savePath3,'time_record','FN_record','FP_record','FN' ,'FP','Sens', 'Spec');  
%     FN = 0;   %  False-negative rate
%     FP = 0;   %  False-positive rate
%     Sens = 0; %  1-FN
%     Spec = 0; %  1-FP
%     for(j=1:positive_num)
%         if(minNum(1,j)~=1)
%             FP = FP+1;
%         end
%     end
%     for(j= (positive_num+1):sum_num)
%         if(minNum(1,j)==1)
%             FN = FN+1;
%         end
%     end
%     FN = FN/negative_num;
%     FP = FP/positive_num;
%     Sens = 1-FN;
%     Spec = 1-FP;
%     FN_record(i) = FN;
%     FP_record(i) = FP;

						