Least squares fitting Gaussian type curve (Q2570711): Difference between revisions

First the authors give necessary and sufficient conditions which guarantee the existence of the least squares estimate of the Gaussian type function \(f(t;b,c,\mu )=be^{c(t-\mu )^2}\). Next they propose the iterative minimization of the weighted functional \(\sum_{i=1}^m p_i(be^{c(t_i-\mu )^2}-y_i)^2\) starting from initial approximation \(f(t_0;b_0,c_0,\mu _0)\) chosen from the experimental data to be sufficiently close to the exact curve. The efficiency of numerical experiments greatly depends on this initial choice.