This repository was archived by the owner on Mar 17, 2023. It is now read-only.
This repository was archived by the owner on Mar 17, 2023. It is now read-only.
A lightweight implementation that dramatically improves runtime of xicor #17
Description
Hello, I wrote a lightweight (and much faster) implementation of the xi correlation metric.
In my tests, this implementation is > 300x faster than the current implementation.
import numpy as np
def _xicorr(X,Y):
n = X.size
xi = np.argsort(X,kind='quicksort')
Y = Y[xi]
_,b,c = np.unique(Y,return_counts=True,return_inverse=True)
r = np.cumsum(c)[b]
_,b,c = np.unique(-Y,return_counts=True,return_inverse=True)
l = np.cumsum(c)[b]
return 1 - n*np.abs(np.diff(r)).sum() / (2*(l*(n-l)).sum())Hopefully this is helpful!