DeWritify

DeDigits

Authors: Mehryar Maalem, Geordie Parappilly

What is the problem?

Inspired by Detextify and one of the most beloved pattern recognition algorithm, K-Nearest neighbours, we decided to implement the algorithm and run it on hand written digits. Usually, the performance of these implementations are very important and we set out to find out whether Haskell is a good language to use for machine learning algorithm.

What is something extra?

- Implementing the relevant model for hand-writing recognition in Haskell.

- We used implementation in F# and python to run and compare the runtime with our Haskell implementation.

What did we learn from this?

- We saw that Haskell does quite well compared to other two languages that we used

- The -threaded feature of the the ghc compiler seems to be doing an amazing job in terms of optimization

- The implementation in Haskell turned out quite elegant yet hard to write.