I am programming a physics simulation (Ising model, if it matters) and I need to fit an 1D decaying exponential function to a set of data and have an estimation of the parameters of the function but however I have searched I cannot for the life of me understand the different fitting methods and choose one that can accomplish this seemingly non-complex task. I have a program in Python that uses scipy.optimization.curve_fit for this function but due to the simulation requiring enormous computing time I am trying to switch to C++.
1 Answer
scipy.optimize.curve_fit uses the lm method by default for unconstrained problems. The documentation for scipy.optimize.least_squares says that
Method ‘lm’ (Levenberg-Marquardt) calls a wrapper over least-squares algorithms implemented in MINPACK (lmder, lmdif).
MINPACK is a Fortran library, but there seems to be a C/C++ version as well: C/C++ Minpack.
Then there are some projects mentioned in the Eigen documentation, foremost the Ceres Solver.
If C is an option, the GNU Scientific Library (GSL) provides linear and non-linear least squares functionality.
