Python (problems with scipy.optimize.curve_fit)

Question

so I'm having trouble with this piece of code here:

import numpy as np
import matplotlib.pyplot as plt
import scipy as sy
import pylab as plb

def oscDecay(x, A, B, C, tau, omega):
    return A*(1+B*np.cos(omega*x))*np.exp(-1*x**2/(2*tau**2))+C


def LMfit(func, x, y, p0, sig):
    #Fits data to non-linear curve using Levenberg-Marquart Method
    #Inputs: func = the function you are fitting data to
    #        x, y = data set
    #       p0 = tuple containing inital guesses for fitting  parameters
    #       sig = uncertainty in values
    #Outputs: nlfit = array containing optimal values for fitting parameters
    #       nlcov = two dimensional array (square root of diagonals contain
    #               uncertainty in fitting parameters)
    nlfit, nlpcov = sy.optimize.curve_fit(func, x, y, p0, sig)
    return nlfit, nlpcov

data=np.loadtxt('testing.txt', skiprows=4)

x=data[:, 0]
y=data[:, 1]
sig=data[:, 2]

#intial parameters
A0=16.5
B0=0.57
C0=17
tau0=30
omega0=7
p0=(A0, B0, C0, tau0, omega0)

nlfit, nlpcov = LMfit(oscDecay, x, y, p0, sig)

When I try to run it, I get this error message:

nlfit, nlpcov = sy.optimize.curve_fit(oscDecay, x, y, p0, sig). AttributeError: 'module' object has no attribute 'optimize'

I'm not sure what it means since my software provided me with scipy.optimize already to use.

You need to do import scipy.optimize, not just import scipy. — BrenBarn
– BrenBarn, Commented May 9, 2016 at 4:29

roadrunner66 · Accepted Answer · 2016-05-09 04:35:27Z

1

In your case the import seems to not have worked as noted by the commenter. I had to make a synthetic sample to test your code. I just left out the sigmas and it worked all right (using Ipython notebooks).

import numpy as np
import matplotlib.pyplot as p
from scipy import optimize as opt
%matplotlib inline

def oscDecay(x, A, B, C, tau, omega):
    return A*(1+B*np.cos(omega*x))*np.exp(-1*x**2/(2*tau**2))+C

# make an example of data with some experimental noise
def oscDecayexp(x, A, B, C, tau, omega):
    rand=np.random.rand(len(x))
    return A*(1+(B+rand/10)*np.cos(omega*x))*np.exp(-1*x**2/(2*tau**2))+C


def LMfit(func, x, y, p0 ,sig=None):
    #Fits data to non-linear curve using Levenberg-Marquart Method
    #Inputs: func = the function you are fitting data to
    #        x, y = data set
    #       p0 = tuple containing inital guesses for fitting  parameters
    #       sig = uncertainty in values
    #Outputs: nlfit = array containing optimal values for fitting parameters
    #       nlcov = two dimensional array (square root of diagonals contain
    #               uncertainty in fitting parameters)
    nlfit, nlpcov = opt.curve_fit(func, x, y, p0, sig)
    return nlfit, nlpcov


#intial parameters
A0=16.5
B0=0.57
C0=17
tau0=30
omega0=7

x=np.arange(0,80,0.02)
y=oscDecayexp(x,A0,B0,C0,tau0,omega0)

p.figure(figsize=(12,8))

p.plot(x, oscDecay(x,A0,B0,C0,tau0,omega0)) # blue
p.plot(x, oscDecayexp(x,A0,B0,C0,tau0,omega0),lw=0.5) #green

p0=(A0, B0, C0, tau0, omega0)
# scipy.optimize.curve_fit(f, xdata, ydata, p0=None, sigma=None, absolute_sigma=False, check_finite=True, **kw)  
nlfit, nlpcov = LMfit(oscDecay, x, y, p0 )

print nlfit

Output:

answered May 9, 2016 at 4:35

roadrunner66

7,9914 gold badges34 silver badges39 bronze badges

Sign up to request clarification or add additional context in comments.

2 Comments

Joe Over a year ago

from scipy import optimize as opt does the trick. upvoted :)

JeeyCi Over a year ago

@ Joe: +1 as well helped me, though it is a strange behavior, but exactly such a trick helped in simple example... though no problems at colab.research.google.com/notebooks/welcome.ipynb at all with such a code

Collectives™ on Stack Overflow

Python (problems with scipy.optimize.curve_fit)

1 Answer 1

2 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

2 Comments

Your Answer

Sign up or log in

Post as a guest

Related