1202.0988 (Massimo Di Pierro)

Improving non-linear fits    [PDF]

Massimo Di Pierro

In this article we propose a method for non-linear fitting that incorporates
some of the features of linear least squares into a general minimum $\chi^2$
fit. Given a fitting function $f(x)$ depending linearly on parameters $\{a_i\}$
and non-linearly on the parameters $\{b_i\}$, the proposed method allows to
reduce the space of the fitting parameters to the $\{b_i\}$ only by determining
the $\{a_i\}$ exactly at each iteration of the non-linear optimizer. We provide
working code and we performed tests of the algorithm using simulated data. We
also show how to include Bayesian priors in the fit to further stabilize it.
Our method is suitable, for example, for fitting with sums of exponentials as
often needed in Lattice Quantum Chromodynamics.

View original: http://arxiv.org/abs/1202.0988