Turns out function tolerance is compared with norm(fval)^2and the default tolerance was 1e-3. Simply changed it to 1e-10 (in accordance with my need) and its working well.
Issue with fsolve tolerance
13 ビュー (過去 30 日間)
I am using fsolve to solve an overdetermined system of equations using the levenberg marquadt algorithm.
It works fine although giving the message "Equation solved, fsolve stalled." This is fine for 90% of my cases.
However, for particular systems of equations, my function value that is returned (fval) after saying that the equation is solved (with fsolve stalled) is
which is unacceptable for my need.
Since the default function tolerance is 1e-6, why does it accept this as a solution?