Something funny going on about cloudtogrid

[kylemcdonald]

thanks! i suspected something like this was happening but never got too deep into it.

before the core of the JV algorithm kicks in, the cost matrices go through an initial set of "reduction" steps to convert them to a form that is easier to work with. if the hungarian algorithm already did some complex reductions (which i think it did) then the JV algorithm can run faster.

i changed the notebook to use np.copy to avoid this mistake.

but if you really want better performance out of the JV algorithm, make sure to cast to integers first and you should get back most of that speed. sorry for misleading!

closed by 8993036