function [parameters,e]=learnrbf(P,T,t)

%本函數完成RBF網絡的參數學習過程，過程中使用具有量子行為的粒子群算法（QPSO）

[m,n]=size(P);
maxsize=1;
popsize=100;
irangew_l=-5;
irangew_r=5;
para=unifrnd(irangew_l,irangew_r,popsize,m);
pbest=para;
MSET=zeros(1,popsize);
MSE=zeros(1,maxsize);


for i=1:popsize
   MSET(i)=mse(P,T,pbest(i,:),t);           %mse()為計算網絡在訓練集上產生均方誤差的函數
end
g=find(MSET==min(MSET));
gbest=pbest(g,:);
mbest=sum(pbest)/popsize;
    
for time=1:maxsize;
    b=1.0-(time-1)/(maxsize-1)*0.5;
    for i=1:popsize;
        for d=1:m;
        a=unifrnd(0,1);
        p=a*pbest(i,d)+(1-a)*pbest(g,d);
        u=unifrnd(0,1);
        if unifrnd(0,1)>=0.5
            para(i,d)=p-b*abs(mbest(d)-para(i,d))*log(1/u);
        else
            para(i,d)=p+b*abs(mbest(d)-para(i,d))*log(1/u);  
        end
        end 
        if mse(P,T,para(i,:),t)<mse(P,T,pbest(i,:),t);
            pbest(i,:)=para(i,:);
        end  
    end
    
    for i=1:popsize
        MSET(i)=mse(P,T,pbest(i,:),t);           
    end
    g=find(MSET==min(MSET));
    if mse(P,T,pbest(g,:),t)<mse(P,T,gbest,t)
    gbest=pbest(g,:);
    end
    mbest=sum(pbest)/popsize;
    MSE(time)=mse(P,T,gbest,t);
end

parameters=gbest;
e=MSE;