The computation of pairwise distances is the main bottleneck of the naive algorithm for distance covariance. Currently we use scipy's cdist for Numpy arrays, and a broadcasting computation in other case.
Any performance improvement to this function is thus well received.
The computation of pairwise distances is the main bottleneck of the naive algorithm for distance covariance. Currently we use scipy's cdist for Numpy arrays, and a broadcasting computation in other case.
Any performance improvement to this function is thus well received.