[SciPy-user] suggestions for improving scipy.integrate.odeint
Hans Fangohr
H.FANGOHR at soton.ac.uk
Tue Apr 19 10:56:31 CDT 2005
Dear all, dear Travis (author of odeint?),
let me first say that scipy is a very useful extension to Python and that
the combination of Python, scipy (and Numerix) and -- in our case --
pylab, rivals MATLAB for our purposes.
Over the last few months, I have been using scipy and in particular
scipy.integrate.odeint in a introductory lecture course on numeric
computation for some engineering degrees.
Thanks to the creativity of the students (they have the most fantastic
ideas you could not dream of) I have come across two minor issues with
odeint, which I'd like to point out here. If these could be resolved (in
the long term), then this would make odeint much easier to use for
beginners. I'll explain why as I go along.
1. The initial value (array) can change while odeint carries out the
integration. This source code
#----------------------------
import Numeric
from scipy.integrate import odeint
def rhs(y,t):
return -y[0]
y0 = Numeric.array([1.0])
print "Initial value:",y0
ts = Numeric.arange(0,10,1)
ys = odeint( rhs, y0, ts )
print "Initial value:",y0
#----------------------------
produces the following output:
#===========================
Initial value: [ 1.]
Initial value: [ 0.00012341]
#===========================
I can kind-of-see why this may happen (without digging in the source).
However, this is highly confusing to the students because y0 is an input
argument to the odeint function and should not change when odeint is
called. (There is, of course, the issue of passing mutable objects such as
lists to functions but I'd rather not start explaining that to
non-computer science students in their very first lectures on Python.
Also, if y0 is a list (rather than an array), this problems does not
exist: if y0 is a list, then it is not modified by running odeint!)
2. The sequence of values t for which odeint returns the numerical
solution y(t) cannot be a list. Here is an example:
#----------------------------
import Numeric
from scipy.integrate import odeint
def rhs(y,t):
return -y[0]
y0 = Numeric.array([1.0])
print "this works"
ts = Numeric.arange(0,10,1)
ys = odeint( rhs, y0, ts )
print "this doesn't"
ts = range(10)
ys = odeint( rhs, y0, ts )
#----------------------------
The output of this code is:
#===========================
this works
this doesn't
Traceback (most recent call last):
File "<stdin>", line 15, in ?
File "/usr/lib/python2.3/site-packages/scipy/integrate/odepack.py", line
114, in odeint
t = t.copy()
AttributeError: 'list' object has no attribute 'copy'
#===========================
Since y0 can be a list, it would be consistent to accept ts to be a list,
too. (This is in fact very useful: by posing problems carefully, students
can use odeint 'from lecture one' even if they haven't learned much about
data types yet.)
I hope this is useful information.
Thanks again for providing such a nice wrapper around ODEPACK.
Cheers,
Hans
More information about the SciPy-user
mailing list