Closed kahaaga closed 1 year ago
@Datseris Ok, I've modified the code so that it uses SVector
s in relevant places. The code allocates, naturally, much less, and is faster. For remaining optimization: I'm not sure if it's worth it. The steps where we could gain some performance is mainly in the SVD step, but that requires an implementation of a non-allocating svd!
(see also benchmark comment in code).
This PR introduces the Lord et al. (2018) differential Shannon entropy estimator.
References
Lord, W. M., Sun, J., & Bollt, E. M. (2018). Geometric k-nearest neighbor estimation of entropy and mutual information. Chaos: An Interdisciplinary Journal of Nonlinear Science, 28(3), 033114.