search

tBuLi / symfit

Symbolic Fitting; fitting as it should be.
http://symfit.readthedocs.org
MIT License
234 stars 17 forks source link

<lambdifygenerated-1>:2: RuntimeWarning: overflow encountered in exp #373

Open ghost opened 9 months ago

ghost commented 9 months ago

Hi.

I am trying to do a curve fitting for two data sets (sig11-lam1) and (sig22-lam2) simultaneously since they have similar parameters. Both sig11 and sig22 are functions of lam1 and lam2.

I tried to write an script for it like this :

from symfit import parameters, variables, Fit, Model, exp
import numpy as np
sig11e = np.array([0.007238495, 0.034467973, 0.062664953, 0.104778522, 0.155771434, 0.223699358, 0.296058426, 0.402029639, 0.950890584, 2.279527369, 4.079938047, 7.336652019, 10.94126853, 13.02172086, 13.53927407])
sig22e = np.array([0.004357766, 0.02221891, 0.029088113, 0.043979816, 0.132058283, 0.186319261, 0.411808169, 0.715913378, 1.984616059, 3.080248551, 5.472464315, 7.560792707, 9.380019492, 11.17472448, 11.85931825])
lam1e = np.array([0.043208, 0.339967999, 0.471335999, 0.560944, 0.692375999, 0.794935999, 0.911128, 1.032007999, 1.198792, 1.365175999, 1.478463999, 1.619112, 1.762359999, 1.847487999, 1.872296])
lam2e = np.array([0.026567999, 0.221599998, 0.327823999, 0.423647999, 0.673327999, 0.754639999, 0.987504, 1.097208, 1.310032, 1.399063999, 1.540599998, 1.643199999, 1.732247999, 1.827775999, 1.870688])
lam1, lam2, sig11, sig22 = variables('lam1, lam2, sig11, sig22')
a1, a2, a3, a4, a5 = parameters('a1, a2, a3, a4, a5')
model = Model({
    sig11: (-2/(lam1**2*lam2**2)) * (a1 + 2*a2*(lam1**2 + lam2**2 + (1/(lam1**2 * lam2**2))-3)+ 3*a3*(lam1**2 + lam2**2 + (1/(lam1**2 * lam2**2))-3)**2) + 2*a1*lam1**2 + 4*a2*lam1**2*(lam1**2 + lam2**2 + (1/(lam1**2 * lam2**2))-3) + 6*a3*lam1*(lam1**2 + lam2**2 + (1/(lam1**2 * lam2**2))-3)**2 + 2*lam1**2*a4*(lam1**2-1)*exp(a5*(lam1**2-1)),

    sig22: (-2/lam1**2) * (a1 + a2*(lam1**2 + lam2**2 + (1/(lam1**2 * lam2**2))-3) + 3*a3*(lam1**2 + lam2**2 + (1/(lam1**2 * lam2**2))-3)**2) + 2*a1*lam2**2 + 4*a2*lam2**2*(lam1**2 + lam2**2 + (1/(lam1**2 * lam2**2))-3) + 6*a3*lam2**2*(lam1**2 + lam2**2 + (1/(lam1**2 * lam2**2))-3)**2,

})
fit = Fit(model, lam1=lam1e, lam2=lam2e, sig11=sig11e, sig22=sig22e)
fit_result = fit.execute()

Where sig11e, sig22e, lam1e and lam2e are the data which I want to do a curve fitting for and they are considered as variables and a1, a2, a3, a4 and a5 are considered as parameters where a1, a2 and a3 are mutual in both sig11 and sig22 functions.

When I run it, I get the following error:

<lambdifygenerated-1>:2: RuntimeWarning: overflow encountered in exp
  return 2*a1*lam1**2 + 4*a2*lam1**2*(lam1**2 + lam2**2 - 3 + 1/(lam1**2*lam2**2)) + 6*a3*lam1*(lam1**2 + lam2**2 - 3 + 1/(lam1**2*lam2**2))**2 + 2*a4*lam1**2*(lam1**2 - 1)*exp(a5*(lam1**2 - 1)) - 2*(a1 + 2*a2*(lam1**2 + lam2**2 - 3 + 1/(lam1**2*lam2**2)) + 3*a3*(lam1**2 + lam2**2 - 3 + 1/(lam1**2*lam2**2))**2)/(lam1**2*lam2**2)
C:\Users\Alireza\anaconda3\Lib\site-packages\symfit\core\objectives.py:318: RuntimeWarning: overflow encountered in square
  (dep_var_value - dep_data) ** 2 / sigma ** 2

Is there something I am missing here or is it because my functions are somehow long?

Thanks