Michael J. Curry

I mean, Haskell RNGs normally live inside IO because they are inherently stateful, right? That's the tricky part where i don't know what things would look like. Ideally there would...

Thanks, this is very solid! Solves my immediate problem for sure. I still would be inclined to work on an actual well-designed package with different PRNGs, low-discrepancy sequences, etc. Any...

Would also be neat to see the Sinkhorn distance, or entropy-regularized Wasserstein norm.

In the context of this package, is there a sensible way to provide an arbitrary cost function when constructing the norm? Or would that not make sense with the design?...

Also, I've just noticed that the norms seem to work well for numbers as well as vectors. Would it be reasonable to define something like `WassersteinNorm{N