> > > Hi, > sorry if this is not the place to post my problem. > I have an over-determined system of equations > representing a non-linear > system and I thought if I could formulate the > he problem as an > optimization one, then I could use matlab's "fmtool" > to solve it. > > My function is describes as: > > A_1 = sqrt((A_2^2)/4 + (1/(B^2))*(P-O)^2)) + > sqrt(A_2^2)/4 + > (1/(B^2))*(P+O)^2)) > > I have "A_1", "P" and "O", and need to find > (optimize) "A_2" and "B". > > How do I write the objective function and the > constraints (I don't have > any) ? I find the help section in matlab a bit > inadequate and thin. > > Thanks, > > Kamran
If you write your function as A_1^2 = (A_2^2/4+sqrt(A_2^2)/4)+1/B^2*((P-O)^2+(P+O)^2) =: p1 + p2*((P-O)^2+(P+O)^2, with p1=A_2^2/4+sqrt(A_2^2)/4, p2=1/B^2 you see that you get a linear fit in the unknowns p1 and p2 under the restriction that both paramters have to be >= 0. You could first try to fit without restrictions on the paramters (easy). If this does not help, try fmincon.