Finding out eigen values and vectors of huge sparse matrices in python is horribly slower than Matlab

Asked May 18 '19 at 16:40

Active May 18 '19 at 17:10

Viewed 224 times

I am trying to find 100 eigen values and vectors of a huge sparse matrix of size 409600x409600. I am using scipy.sparse.linalg.eigs for this and it's taking ages to find the result whereas eigs on matlab solves it within 10 mins.Any suggestions on how to speed it up?

Python:

eigenvectors, eigenvalues = scipy.sparse.linalg.eigs(Laplacian, k=100, which='SM')

Matlab:

eigCnt = 100; [eigenvectors, eigenvalues] = eigs(Laplacian, eigCnt, 'SM');

where Laplacian is a sparse matrix of size 409600x409600 with 10418204 non zero entries `

edited May 18 '19 at 17:10

asked May 18 '19 at 16:40

S Chakraborty

1

Show us your matlab code solving the problem, also attached the format of the matrix – Adam May 18 '19 at 17:03
1

MATLAB has a lot of corporate dollars to throw at this sort of calculation. But you can read the sparse `eigs` docs as well as we can. – hpaulj May 18 '19 at 17:10
1

@Adam I've updated it – S Chakraborty May 18 '19 at 17:12
@hpaulj I am looking to do it in Python. Hence the question :) – S Chakraborty May 18 '19 at 17:13
1

We get a lot of questions here about MATLAB being faster than Python. The one is an expensive software package, the other is free. You have to pick: cheap or fast? – Cris Luengo May 18 '19 at 18:40

Finding out eigen values and vectors of huge sparse matrices in python is horribly slower than Matlab

0 Answers0