Accompanying code

Paper: Pettersson M., Ekeroth, J. L., and Russo, A. Calculating Function Sensitivity for Synthetic Data Algorithms, ACM 35th Symposium on Implementation and Application of Functional Languages (IFL), 2023.

The improved version of the trick

To try trick', then

cabal repl dpella-isensitivity

and once in the Haskell's prompt:

*DSL.Deep> :module TypeLift.Trick

Prelude TypeLift.Trick> :t foo
foo :: Num p => PatternsOf T '[(), o2, o3] -> p

Prelude TypeLift.Trick> :t trick
trick
  :: Data.SOP.Constraint.All HasBeenMatched ts =>
     (PatternsOf t ts -> r) -> [r]

Prelude TypeLift.Trick> trick foo
[10,1]

To run MWEM

cabal repl dpella-benchmark

and then, from the Haskell's prompt:

*Main> main
producing (1.0e-2,10) ..
producing (0.1,10) ..
producing (1.0,10) ..
ByeBye

The specification of the workload using the improved version of the trick is in the file ./dpella-benchmark/app/Workload.hs. The generated synthetic datasets are in the results folder.

Disclaimer

This code is for demonstration purposes. It is not intended to be used in production.