InfSumPy is a Python package that evaluates infinite positive sums with guaranteed error. Using ratio and integral tests we evaluate series that pass these tests with controlled error.
Make sure you have the mpmath library installed:
pip install mpmath
To install the package, run the following command:
pip install infsumpy
We have the transformations implemented above, and for use, we have the infsum
function.
Which receives from input:
-
A series: In the form of a function f:
$\mathbb{N} \to \mathbb{R}$ . -
Method: Can be
ratio
,integral
,threshold
orfixed
. - Max terms: The maximum number of terms.
- Start terms: The index of the first term of the series.
-
Epsilon (optional): The expected error tolerance (if the method is
ratio
,integral
orthreshold
). -
L (optional): Limit of the ratio of terms (if the method is
ratio
). -
Integral of series (optional): The function of g(n) = ∫_n^∞ f(x) dx for the integral test (if the method is
integral
). -
Precision (optional): The precision for the
mpmath
library (default value is 53).
The function returns the number of terms used in the sum and the approximation.
from infsumpy import infsum
# the infinity sum of n/(2**n) pass in the ratio test with limit L = 1/2,
# then we can evaluate with controled error
print(infsum(lambda n: n/(2**n), 'ratio', max_terms=10**4, initial=1, eps=2**(-52), L=1/2))
> (56, 2.0)
from infsumpy import infsum
# the infinity sum of 1/n**2 pass in the integral test with integral
# g(n) = ∫_n^∞ 1/x**2 dx = 1/n, then we can evaluate with controled error
print(infsum(lambda n: 1/(n**2), 'integral', max_terms=10**4, initial=1, eps=10**(-3), g=lambda n: 1/n))
> (499, 1.64493406229104)
from infsumpy import infsum
# we can also use a stoping criterio such that sum until the n-th are less
# than the epsilon, here for the infinity sum of 2/(2**n)
print(infsum(lambda n: n/(2**n), 'threshold', max_terms=10**4, initial=1, eps=2**(-52)))
> (57, 2.0)
from infsumpy import infsum
# we can just sum a fixed number of terms of the infinite sum of 2/(2**n)
print(infsum(lambda n: n/(2**n), 'fixed', max_terms=10**4, initial=1))
> (10000, 2.0)