Hi,

i want to minimize a cost function, which goes from R^3 to R^1. So far I have attempted to use LevenbergMarquardt, from the NonlinearOptimization package to do the job. However it appears to not like the fact that my value space is smaller than my parameter space. Is it possible to use LM for this kind of problem at all? Am I misunderstanding LM? What alternatives do I have?
I have attempted to use the HybridNonLinearSolver which is included in the same package. However it passes a 3x3 Jacobian, even though the Jacobian of my Function is just 1x3, which causes a run-time error.

Any ideas?

Thanks
Frank

LM solve problems of the form:

cost(x) = sum_i (f_i(x))^2

The numer of f_i must be larger than the dimension of x. If you cannot decompose your cost function this way then you must use a totally different kind of solver.

Hm ok... and the HybridNonLinearSolver works the same way?

So there is nothing in Eigen which can be used to minimize a function from R^3->R?