calc_posprob.m

Graphical User Interface, matlab图形界面程序值得参考欢迎下载

% This file computes the probabilistic outputs of the SVM using the sigmoid
% method by J. Platt
% Used In Conjunction with D2CMatlab
% Written by D.Lai 2006 
% Requires name of input file containing SVM decision values f(x). f(x) can
% be obtained using leave one out, crossvalidation or hold out methods.
%
function calc_posprob(h,handles)

F=load(handles.SVMdecisionfile);
[TotalRows TotalCol]=size(F);
OutputFile=handles.SVMposfile;

%determine number of positive examples and number of negative examples
NUMNEG=0;
NUMPOSITIVE=0;
for i=1:TotalRows
 if (F(i,2)<0)
     NUMNEG=NUMNEG+1;
 end
end
NUMPOSITIVE=TotalRows-NUMNEG;    

%output parameters
A=0;
B=0;

%Initialization
A=0;
B=log10((NUMNEG+1)/(NUMPOSITIVE+1));
hiTarget=(NUMPOSITIVE+1)/(NUMPOSITIVE+2);
loTarget=1/(NUMNEG+2);
lambda=0.001;
olderr=1e300;

pp=zeros(TotalRows,1);
temp=(NUMPOSITIVE+1)/(TotalRows+2);

for i=1:TotalRows
    pp(i,1)=temp;
end

count=0;

%Begin iteration

for it=1:100
    a=0;
    b=0;
    c=0;
    d=0;
    e=0;
    
    %Compute the coefficients of the sum terms, Hessian and gradient terms
    %of the max likelihood function
    for i=1:TotalRows
        if(F(i,2)==1) %if target is positive
           t=hiTarget;
        else
           t=loTarget;     
        end
                      
        d1=pp(i)-t;
        d2=pp(i)*(1-pp(i));
        a=a+F(i,1)*F(i,1)*d2;
        b=b+d2;
        c=c+F(i,1)*d2;
        d=d+F(i,1)*d1;
        e=e+d1;
    end
    
    %Exit if gradient is small
    if ( abs(d)<1e-9 ) & (abs(e) < 1e-9) 
        break;
    end
    
    
    oldA=A;
    oldB=B;
    err=0;
    
    while(1)
     det=(a+lambda)*(b+lambda)-c*c;
     
     if(det==0)
         lambda=lambda*10;
         continue;
     end
     
     %Compute Step
     A=oldA+((b+lambda)*d-c*e)/det;
     B=oldB+((a+lambda)*e-c*d)/det;
     
     %compute goodness of fit
     err=0;
     for i=1:TotalRows
      p=1/(1+exp(A*F(i,1)+B));
      pp(i)=p;
      
      % Log 0 returns -200
      logtemp1=log10(p);
      
      if(p==0)
          logtemp1=-200;
      end          
      logtemp2=log10(1-p);
      
      if(1-p==0)
          logtemp2=-200;
      end         
      err=err-t*logtemp1+(1-t)*logtemp2;
      
     end
      
     if(err<olderr*(1+1e-7))
         lambda=lambda*0.1;
         break;
     end
     
     lambda=lambda*10;
     if(lambda>=1e6)
         break;
     end
          
    end

    diff=err-olderr;
    scale=0.5*(err+olderr+1);
    if((diff > -1e-3*scale) & (diff <1e-7*scale))
      count=count+1;
  else
      count=0;
  end
  
  olderr=err;
         
    if (count==3)
        break;
    end   
   
end 

fid=fopen(OutputFile, 'w+');
for i=1:TotalRows
 fprintf(fid,'%f %f %f \n',F(i,1),F(i,2),pp(i));  
end
fclose(fid)

handles.SVMPP=[F(1:TotalRows,2)  pp];
handles.SVMPP=sortrows(handles.SVMPP,1);
guidata(h, handles);