% anfis_ap1.m (ANFIS training file for XOR )
% ANFIS: Adaptive Neuro-Fuzzy training of Sugeno-type FIS.
data=[...
   0 0 0
   0 1 1
   1 0 1
   1 1 0];  % data = 4*3 矩陣 (本例)
numpts=length(data(:,1)); % 4 (本例)
trndata=data(1:1:numpts,:); % trndata = data
chkdata=data(2:2:numpts,:); % chkdata=[0 1 1; 1 1 0]
figure(1);
plot3(trndata(:,1),trndata(:,2),trndata(:,3),'o',chkdata(:,1),chkdata(:,2),chkdata(:,3),'x');
title('train data and check data before training'); %繪 學習前：訓練點(o)，驗證點(x)
nummfs=2 % 單一變數所使用的 MF 數
mftype='gbellmf'; % 隸屬函數的形式
fismat=genfis1(trndata,nummfs,mftype); % 產生模糊關係矩陣
[x,mf_x]=plotmf(fismat,'input',1);
[y,mf_y]=plotmf(fismat,'input',2);
figure(2);
subplot(311)
plot(x,mf_x,'o');hold on;
plot(y,mf_y,'x');hold on; % 繪 學習前：I/P MFs 
title('The input membership function before training')
%Begin to train the membership function by ANFIS
numepochs=30;   % 值越大，反覆學習次數越多
[fismat1,trnerr,ss,fismat2,chkerr]=anfis(trndata,fismat,numepochs,NaN,chkdata);
                % ANFIS 開始訓練
[x,mf_x]=plotmf(fismat1,'input',1);
[y,mf_y]=plotmf(fismat1,'input',2);
subplot(312)
plot(x,mf_x);hold on;
title('The input1 membership function after training')
subplot(313)
plot(y,mf_y);hold on;    % 繪 學習後：I/P MFs
title('The input2 membership function after training')
%Evaluate the fis matrix
out=evalfis([trndata(:,1) trndata(:,2)],fismat1);
% 輸入data= [trndata(:,1) trndata(:,2)]=[0 0; 0 1; 1 0;1 1];
figure(3);
trndata=data(1:1:numpts,:);
chkdata=data(2:2:numpts,:);
plot3(trndata(:,1),trndata(:,2),trndata(:,3),'o',chkdata(:,1),chkdata(:,2),chkdata(:,3),'x');hold on;
plot3(trndata(:,1),trndata(:,2),out);  %繪 學習後：訓練點(o)，驗證點(x)
title('line represents that train data and check data after training');