I am Adam Ścibior, a third year PhD student at University of Cambridge and Max Planck Institute for Intelligent Systems. This post is about a last minute informal workshop on numerical programming and data science that was held at ICFP this year.
The workshop was put together by Dominic Steinitz and Marco Zocca. It was opened by Dominic who presented a manifesto for doing numerical programming in Haskell. Long story short there seems to be an opportunity here if only we can build enough momentum as a community.
Next I presented my library for probabilistic programming in Haskell named monad-bayes. The library demonstrates a novel technique for building Bayesian inference algorithms in a compositional fasion and should soon be officially released, in the meantime it’s already available on GitHub.
Then Praveen Narayanan presented a probabilistic programming language Hakaru and demonstrated some of its power for symbolic manipulation of probabilistic program. After that Trevor McDonell presented the accelerate package for fast array computation on both CPUs and GPUs.
Afterwards Belazs Komuves talked about how he does not usually find Haskell the right tool for the job in numerical computing and how Matlab has multiple advantages so as a tool it can be very difficult to beat in practice. Next up was Michał Gajda, but unfortunately I was called upon to perform my student volunteer duties in another session so I had to leave the talk early.
All in all it was a great opportunity to bring together a group of people interested in numerical programming in Haskell and hopefully next year it will turn into a proper workshop.