source: prextra/rbsvc.m @ 10

%RBSVC Automaric radial basis SVM
%
%   W = RBSVC(A,FID)

function w = rbsvc(a,fid)

if nargin < 2, fid = []; end
if nargin < 1 | isempty(a)
        w = mapping(mfilename,fid);
        return
end

nu = max(testk(a,1),0.05);

d = sqrt(+distm(a));
sigmax = min(max(d)); % max: smallest furthest neighboe distance

d = d + 1e100*eye(size(a,1));
sigmin = max(min(d)); % min: largest nearest neighbor distance

q = (1/sigmin - 1/sigmax)/9;
sig = [1/sigmax:q:1/sigmin];
sig = ones(1,length(sig))./[1/sigmax:q:1/sigmin];
if isempty(sig)
        sigopt = sigmin;
else
        w = [];
        errmin = inf;
        sigopt = 0;
        prprogress(fid,'     ')
        for j=1:length(sig)
                s = sprintf('\b\b\b\b\b%5.0f',j);
                prprogress(fid,s);
                err = crossval(a,svc_nu([],'r',sig(j),nu),10,1,fid);
                if err < errmin
                        errmin = err;
                        sigopt = sig(j);
                end
        end
        prprogress(fid,'\b\b\b\b\b')
end
w = svc_nu(a,'r',sigopt,nu);