I’d love if someone could explain in applied, non-category-theoretic terms why monads are so attractive theoretically and how they are useful practically. If category theory is the only way they can be explained, they’ll certainly “stay theoretical.” Our users aren’t going to know lambda calculus either, but we are going to add closures for lambdas, so maybe we’ll nudge them a bit in that direction.

]]>In addition, the GPL is essentially only preventing a company from redistributing a closed-source version of RStan. And I have no earthly idea why a company would want to do that.

Such a company can already use RStan internally or host RStan on a cloud without making its Stan models or data public.

]]>Theoretically, there could be a big win in a Haskell monad interface to an API for Stan – certainly the Haskell linear programming monad (or at least the one I have used) is unbelieveably flexible and expressive, but lets face it, that win is likely to stay theoretical, and anyway Stan models are much smaller than linear programming models can be, so the win is likely smaller anyway.

]]>(That’s the Bank of England announcement; as end-users, they care about results a lot more than methods.)

]]>I don’t think we’ll move to an ML-style type system (afraid I’ve never used Haskell), but we will be adding functional programming through simple well-founded typing where (A -> B) is a type if A and B are types. One of the things we’ll introduce before a proper type system is a parallel map function with signature (A -> B x B[]) -> A[].

]]>I didn’t see any references to Stan in that document, but then it didn’t go over other tools, either.

]]>Incidentally, a Stan model will be the basis of (one component of) an anomaly detection tool developed for use by the Bank of England. The guy who gritted his teeth and undertook to put it a Docker container and get our continuous integration tools to work with it was not happy with me, even afterI told him he’d done out-Stan-ding work; somehow the pun didn’t improve his outlook…

http://www.bankofengland.co.uk/Documents/fintech/mindbridgeaipoc.pdf

]]>