Scipy correlation function slow

Question

I compared different convolution / correlation methods of two signals using numpy / scipy. It turns out that there are huge differences in speed. I compared the following methods:

• correlate from numpy package (np.correlate in plot)
• match from scipy.signal package (sps.correlate in the plot)
• fftconvolve from scipy.signal (sps.fftconvolve in the plot)

Now, of course, I understand that there is a significant difference between fftconvolve and two other functions. I do not understand why sps.correlate is much slower than np.correlate. Does anyone know why Scipi uses an implementation that is much slower?

For completeness, here is the code that creates the plot:

import time import numpy as np import scipy.signal as sps from matplotlib import pyplot as plt if __name__ == '__main__': a = 10**(np.arange(10)/2) print(a) results = {} results['np.correlate'] = np.zeros(len(a)) results['sps.correlate'] = np.zeros(len(a)) results['sps.fftconvolve'] = np.zeros(len(a)) ii = 0 for length in a: sig = np.random.rand(length) t0 = time.clock() for jj in range(3): np.correlate(sig, sig, 'full') t1 = time.clock() elapsed = (t1-t0)/3 results['np.correlate'][ii] = elapsed t0 = time.clock() for jj in range(3): sps.correlate(sig, sig, 'full') t1 = time.clock() elapsed = (t1-t0)/3 results['sps.correlate'][ii] = elapsed t0 = time.clock() for jj in range(3): sps.fftconvolve(sig, sig, 'full') t1 = time.clock() elapsed = (t1-t0)/3 results['sps.fftconvolve'][ii] = elapsed ii += 1 ax = plt.figure() plt.loglog(a, results['np.correlate'], label='np.correlate') plt.loglog(a, results['sps.correlate'], label='sps.correlate') plt.loglog(a, results['sps.fftconvolve'], label='sps.fftconvolve') plt.xlabel('Signal length') plt.ylabel('Elapsed time in seconds') plt.legend() plt.grid() plt.show() 

Answer 1

According to the documentation, numpy.correlate was designed for 1D arrays, and scipy.correlate can accept ND arrays.

A transparent implementation, which is more general and therefore complex, apparently carries additional computational overhead. You can compare C code between numpy and scipy implementations.

Another difference may be, for example, that the numpy implementation becomes better than the vectorized compiler on modern processors, etc.