Erich Schubert

**Distance matrix:** Indeed, the distance computations are the main cost, but that is also the baseline any non-quadratic method will need to beat. But even pairwise distance computations can be...

It does increase the run time with the quadratic cost of the similarity matrix though.

For this kind of code, pure python is unsuitable - the python interpreter is too slow, the boxing overheads are too large. You need to use at least Cython. The...

High on my font wish list are sans serif fonts, such as newtxsf which plays nicely with the popular Fira Sans font.

Yes, sans serif obviously. Fira Sans is pretty popular, and I'd like math typography consistent with Fira Sans. Fira Math seems to be work in progress. newtxsf supposedly is quite...

A collection of standard benchmark data sets for FIM is here: http://fimi.uantwerpen.be/data/ An effort to benchmark these implementations should likely begin with this widely used collection.

The performance of MLXtend "apriori" is pretty poor. It does **not** implement the optimizations of the original Apriori algorithm such as the prefix join to avoid generating unnecessary candidates or...

@pintergreg the actions often (e.g. when generated by pdflatex?) have a `action["/D"]` value that is the destination name. But I haven't found a way to resolve this name yet...

A code snipped working for me, but likely not at all robust (built using hints from #173): ```python # Note: I want page numbers starting at 1 instead of 0...

BUILD and SWAP, as in the PAM paper. > Kaufman, L. and Rousseeuw, P.J. (1987), Clustering by means of Medoids, in Statistical Data Analysis Based on the L1–Norm and Related...