find_fit piecewise defined functions

Hi all! I'm working with the function find_fit and now I need to fit a piecewise function, but I have tried it returns errors. How could I fit a picewise function?

Thanks

edit retag close merge delete

2

Yes. It turns out that piecewise does not have a member function variables. So len(variables) retruns a error. Is there a ticket about this? I could not find any. If the function happens to be piecewise we should call length and not len(variables)

( 2012-01-14 11:19:28 -0500 )edit

Sounds like you should open a ticket for this :)

( 2012-01-16 01:55:38 -0500 )edit

A ticket has been created (http://trac.sagemath.org/sage_trac/ticket/12316). I will try to fix it soon.

( 2012-01-16 08:48:03 -0500 )edit

Ok, thank you.

( 2012-01-16 09:07:24 -0500 )edit

Sort by » oldest newest most voted

According to the documentation of find_fit, the "model" can be a plain python function:

model  Either a symbolic expression, symbolic function, or a Python function. model has to be a function of the variables (x1;x2;:::;xk) and free parameters (a1;a2;:::;al).

and nothing stops you from using conditionals inside f:

def f(t,a) :
if t<1/2:
return t*a
return (1-t)*a

data = [(i,i*(1-i)) for i in xsrange(0, 1, 0.1)]

var('a, x')
d  = find_fit(data,f, parameters=[a], variables=[x], solution_dict=True)
a0 = d[a]
print a0
point(data) + plot(lambda t:f(t,a0), 0,1)

more

yes that is true, but i think conditional statement is slower.

( 2012-01-17 13:49:43 -0500 )edit

If it is slower than Piecewise, I'd say that's only because of fast_callable. But you can also compile the function f with cython

( 2012-01-17 21:48:02 -0500 )edit