CODEWITHSUNDEEP
pythonscipyscipy-optimize
Jostein Trondal
Jostein Trondal

Reputation: 105

How to dynamically specify parameters when using scipy.optimize.curve_fit

I have a complicated model function that is composed by 12 other functions. I have about 10 parameters. I use the model function both to fit data and to generate contour plots. Sometimes I am only interested in estimating only a few parameters, and fixing the rest, setting them as global variables. Other times I want to estimate most parameters, and fix only a few. Every time I change the set of which parameters to estimate and which to fit, I go through my 13 functions, changing all the calls to only include those parameters I want to estimate. Can this be done more dynamically? Perhaps also avoiding global variables?

Upvotes: 0

Views: 505

Answers (1)

blunova
blunova

Reputation: 2532

The partial function from the functools library can help you (here is the documentation).

For instance, one of your functions has the following form:

def f(x, a, b, c, d):
    # a very complex expression
    return ...

where a, b, c and d are the parameters to be optimized using curve_fit by SciPy.

Suppose that you need to optimize the parameters a and b and you want to keep c and d fixed. You can write:

# constant value
c0 = ...
# constant value
d0 = ...
f_partial = partial(f, c=c0, d=d0)

now you get a new function with a simplified signature and you can pass f_partial to curve_fit and only a and b will be optimized.

You can now leverage the partial function to choose which parameters to optimize.

This tutorial can also help you to understand better how the partial function works.

Upvotes: 1

Related Questions