[SciPy-user] optimization question

[Lin Shao]

Can't you express c(x) and d(x) as functions in terms of x? If so,
then I assume you'll get a bunch of scalar parameters c1,c2... for
c(x) and d1,d2,... for d(x). Then what you need is to fit
(a,b,c1,c2...,d1,d2...) using your measured data and known x_i.

scipy.optimize.leastsq is a good choice. But for some reason, the
Jacobian matrix has to be "column major" (each column represents one
observation point) and set col_deriv to 1. Otherwise it won't work. I
reported this bug weeks ago, but nobody seems to have answered with
authority that if it's a bug or not.