1. 4
  1. 1

    I like this abstract, printed it out to read. Anyone have other papers in this realm that they think are cool?

    1. 2

      I like “Compiling to Categories” by Conal Eliott. His original goal was to compile Haskell functions to hardware, like FPGAs. But instead of creating a whole separate compiler, he added a feature to GHC that lets you “Overload lambda”. That means you can take an ordinary lambda expression like (\a b -> square a + square b) and interpret it as a combinational circuit, or a GPU shader, or an SVG of the dataflow… One of the many applications is automatic differentiation.

      I also like “Dex: array programming with typed indices”:

      The Matlab/NumPy model of array programming avoids explicit loops and indexing and it can be impressively concise. But it can also feel obfuscated. […]

      The traditional scalar loops of Fortran and C are more explicit, but much heavier. […]

      Can we have the best of both? The clarity of indexing explicitly and the succinctness of looping implicitly? Perhaps!

      When I used Matlab for a class, I found it really hard to keep track of the implicit array indexing. It’s nice to know I’m not the only one, and their solution looks really slick!