Open andeElliott opened 6 years ago
@martintoreilly speedupRV2 code has been commented therefore is ready for you to read through.
There seems to be a difference between the results on Linux and Mac, with linux resulting in incorrect answers with the C based methods (the R only methods are giving reasonable answers)
There is also a problem with the ver2 method in small graph sizes
Exploring timings and comparing a BA (varying network sizes with fixed degree) with ER seems to take around the same amount of time as the optimise method with the cpp version of EMD
Note, we should do extensive testing before we merge the faster version to make sure they are correct. @martintoreilly p I have seen some differences that seems larger than i would expect by machine precision.