Hello! I try to fit a non linear model. It is a quite complex (4 parameters) biological model and therefore the optimization surface is probably very rough, i.e. there are several optima. The optimum being found by the algorithm depends thus on the starting values I feed in. I have seen the "SANN" method in the optim function which is probably a good solution. My problem now is that I have to constrain my parameters (from biological point of view, 3 pars >0, 1 par>1, maybe also set upper boundaries) but if I do the "SANN", where no boundaries can be set, in only about 20% of the cases a solution is found that satisfies the biological constraints. Is there any way in R to perform a constrained simulated annealing? Thanks a lot Fabian [[alternative HTML version deleted]]