examp.m

这是在网上下的一个东东

%
% Reinitialize everything.
%
clear
rand('seed',0);
randn('seed',0);
%
% Get the problem.
%
load shawexamp.mat
%
% Set the data covariance matrix.
%
CD=1.0e-12*eye(20);
%
% Setup the prior.
%
mprior=0.5*ones(20,1);
CM=0.25*eye(20); 
%
% Compute the posterior distribution.
%
[covmp,mmap]=bayes(G,mprior,CM,dn,CD);
%
% Compute a second solution with a more diffuse prior.
%
mprior2=mprior;
CM2=100*CM;
[covmp2,mmap2]=bayes(G,mprior2,CM2,dn,CD);
%
% Now, produce some plots.
%
figure(1);
clf;
bookfonts;
plotconst(spike,-pi/2,pi/2);
axis([-2 2 -0.5 1.5]);
hold on
plotconstc(mmap,-pi/2,pi/2,'k--');
xlabel('\theta');
ylabel('Intensity');
legend('Target Model','MAP Solution');
%print -deps mmap.eps
%
% Now, with error bars.
%
figure(2);
bookfonts;
plotconst(mmap,-pi/2,pi/2);
axis([-2 2 -1.0 1.5]);
hold on
plotconstc(mmap+1.96*sqrt(diag(covmp)),-pi/2,pi/2,'k--');
plotconstc(mmap-1.96*sqrt(diag(covmp)),-pi/2,pi/2,'k--');
xlabel('\theta');
ylabel('Intensity');
legend('MAP Solution','95% Prob Interval');
%print -deps mmapb.eps
%
% Generate a random solution.
%
figure(3);
bookfonts;
mmapsims=simmvn(mmap,covmp);
plotconst(mmapsims,-pi/2,pi/2);
bookfonts;
xlabel('\theta');
ylabel('Intensity');
%print -deps mmapsims.eps
%
% Now, with error bars for the more diffuse prior.
%
figure(4);
clf;
bookfonts;
plotconst(mmap2,-pi/2,pi/2);
hold on
plotconstc(mmap2+1.96*sqrt(diag(covmp2)),-pi/2,pi/2,'k--');
plotconstc(mmap2-1.96*sqrt(diag(covmp2)),-pi/2,pi/2,'k--');
xlabel('\theta');
ylabel('Intensity');
legend('MAP Solution','95% Prob Interval');
%print -deps mmap2.eps