Skip to main content

Pairwise Summation algorithm implementation with python

Pairwise summation algorithm is a summation algorithm like Kahan Algorithm but does perform better when compared to the Kahan Algorithm. This is because of the Divide and Conquer approach followed by this algorithm. According to Wikipedia, this is the official summation algorithm used in Numpy.


 If you don't know what divide and conquer is all about, then you should see the following video: lecture 3: Divide and Conquer by MITOCW.

 To be simple, this algorithm takes a list of numbers whose length is at least greater than 0 and also a sufficiently small number N < n(length of the list).

 Now the list is broken down into sublists of length N, we find the sum of these sublists and finally add all the summations together to get the result. As simple as that.

 Up next is the program for this algorithm which I have written using python. But if you want further reference then you can see the Wikipedia pairwise summation algorithm. Remember that you can only understand the algorithm better when you will take an example. For example, you can plug in [1, 2, 3] as the list. Take a paper and write down the output of each and every statement.

Program

Program is self explanative and I don't think this one needs a detailed explanation.

 Let us test if the algorithm is performing as expected. I am just testing if there is precision or not. You can also use timeit to check the speed of the algorithm 
>>> aList = [0.1]*10
>>> aList
[0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1]
>>> pairwise(aList)
1.0
>>> pairwise(aList) == 1.0
True
>>>
As you can see the algorithm performs as expected.  But remember that the condition N < n is taken for granted. If you change the value of N to be less than the length of the list then you will get an error.

You can replace the lines 10-12 with sum function.

But guys I don't know why, this algorithm is giving a wrong result when the input is [0.1, 0.1, 0.1]. Remaining results are satisfactory, at-least the test cases which I have done! If you are aware of why this is happening, then please, please comment in the comment box.

If you have any doubt or haven't understood anything, then comment in the comment box below and I will be glad to help you.

 If you have any suggestions or feedback or if you have found a typo then please comment in the comment box below and I will be very happy to see your post.

 You can also contact me.

 Thank you. Have a nice day😃.
Labels:

Popular posts from this blog

Project Euler Problem 67 Solution with Python

Maximum path sum II By starting at the top of the triangle below and moving to adjacent numbers on the row below, the maximum total from top to bottom is 23. 3 7 4 2 4 6 8 5 9 3 That is, 3 + 7 + 4 + 9 = 23. Find the maximum total from top to bottom in triangle.txt (right click and 'Save Link/Target As...'), a 15K text file containing a triangle with one-hundred rows.
Read more

Problem 60 Project Euler Solution with python

Prime pair sets The primes 3, 7, 109, and 673, are quite remarkable. By taking any two primes and concatenating them in any order the result will always be prime. For example, taking 7 and 109, both 7109 and 1097 are prime. The sum of these four primes, 792, represents the lowest sum for a set of four primes with this property. Find the lowest sum for a set of five primes for which any two primes concatenate to produce another prime. This problem is j u st a brute force problem. If you have come here because you don't know the limit upto which you will h ave to gener ate the prime numbers t hen go ahe ad and t r y with 10,000 . When I first start ed solving the problem I chose 1 million(beca use most of the problem s on project E uler have this limit ), but it took very long for the computer to fin d the solution. After searching on the internet then I found many people choosing 10, 000 so I have changed my in put f rom 1 million to 10000 and the output was f ast. He...
Read more

Project Euler Problem 66 Solution with python

Diophantine equation ¶ Consider quadratic Diophantine equations of the form: $$ x^{2} – Dy^{2} = 1 $$ For example, when $D = 13$, the minimal solution in $ x $ is $ 649^{2} – 13 \times 180^{2} = 1 $ It can be assumed that there are no solutions in positive integers when $ D $ is square. By finding minimal solutions in $ x $ for $ D = {2, 3, 5, 6, 7} $, we obtain the following: $$ 3^{2} – 2×2^{2} = 1 $$ $$ 2^{2} – 3×1^{2} = 1 $$ $$ 9^{2} – 5×4^{2} = 1 $$ $$ 5^{2} – 6×2^{2} = 1 $$ $$ 8^{2} – 7×3^{2} = 1 $$ Hence, by considering minimal solutions in $ x $ for $ D ≤ 7 $, the largest $ x $ is obtained when $ D = 5 $. Find the value of $ D ≤ 1000 $ in minimal solutions of $ x $ for which the largest value of $ x $ is obtained.
Read more