Module project.test.test_matrix

import numpy as np
from scipy import sparse, stats


def generate_smatrix(m, n, density=1, bound=0.5, **kwargs):
    return sparse.random(m,
                         n,
                         density=density,
                         format="csr",
                         data_rvs=stats.uniform(-bound, 1).rvs)


def scale_spectral_smatrix(matrix: sparse.issparse,
                           spectral_radius=1.25,
                           in_place=False):
    try:
        eigs = sparse.linalg.eigs(matrix)[0]
    except sparse.linalg.ArpackNoConvergence as e:
        eigs = e.eigenvalues
    finally:
        spectral = (spectral_radius / max(abs(eigs)))

    return matrix * spectral


if __name__ == "__main__":
    from timeit import default_timer as timer
    from pprint import pprint
    tempos = [(timer(),
               scale_spectral_smatrix(generate_smatrix(2000, 2000,
                                                       0.0001)), timer())
              for _ in range(1000)]
    stat = np.mean([z - x for (x, _, z) in tempos])
    pprint(stat)