The Hasktorch library makes available hundreds of mathematical operations including linear algebra, GPU computation, non-linear neural network transformations and their gradients, high level lapack/blas interfaces, and probability distributions/sampling tools. These comprehensive operations provide a foundation for exploring machine learning systems with statically typed functional programming. It builds on top of C libraries that have been in development for over a decade and are the backend for the PyTorch library.
Modern machine learning frameworks generally target imperative, weakly-typed languages such as Python, Lua, and R, and inherit the design of their host language. Hasktorch's higher-level APIs are designed with several basic principles that, while well-known in functional programming, are not commonly incorporated together with machine learning. We believe that giving developers access to simple Haskell bindings can help create a new wave of more expressive and safer machine learning models.
Sam Stites is a Haskell engineer and machine learning researcher at Sentenai.
The meetup will be at the Thoughtbot location in downtown Boston. Food should be available near the 6:30PM start time, provided by Thoughtbot.