HERE IS MY PROGRAM:
gamma = linspace(0,0.5,10)
l2norm = zeros(size(gamma));
l1norm = zeros(size(gamma));
for k = 1:length(gamma)
fprintf(1, ‘%8.4e’, gamma(k));
cvx_begin
variable I(3000)
minimize(norm(Output’-S’*I)+gamma(k)*norm(I,2))
cvx_end
l1norm(k) = norm(I,1);
l2norm(k) = norm(Output’-S’*I);
fprintf(1, ’ %8.4e %8.4e\n’, l1norm(k), l2norm(k));
end
THE BLUE ONE IS THE OVERFITTING CURVE
As the script shows, which is similar to the example in the guide book,an optimal trade-off curve, but after adding the ‘gamma’ regularization, little effects occured on my figure. It’s still overfitting, besides, the gamma is limited from 0 to 0.3, or there will be no fitting curves in the figure. How can i solve this??