Digit Recognizer solutions from the F# machine learning coding with @brandewinder
Haskell C# F# Standard ML Rust Elixir Other
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
bryan_hunter+elixir
drewr+haskell
jorendorff+haskell
jorendorff+rust
kimsk+fsharp
luketopia+fsharp
luketopia+scala
mark_wutka+haskell
mark_wutka+ocaml
mark_wutka+sml
ralmlopez-fsharp
.gitignore
README.md
test-sample.csv
training-sample.csv

README.md

Kaggle's digit recognizer challenge

This is NashFP's playground for solving Kaggle's Digit Recognizer challenge.

In August 2013, Mathias Brandewinder visited Nashville to lead a machine learning in F# dojo. The dojo built on top of the digit recognizer challenge. It was a blast. Later this dojo was packaged (with slides, data, and solutions) and placed here.

The summary for the dojo:

This Dojo is inspired by the Kaggle Digit Recognizer contest. The goal is to write a Machine Learning classifier from scratch, a program that will recognize hand-written digits automatically. It is a guided Dojo, suitable for beginners: the Dojo comes with a Script file, with specific tasks to complete, introducing along the way numerous F# concepts and syntax examples. The Dojo is on the longer side; for a group with limited F# experience, it is recommended to schedule it for 2 hours and a half.

Playground

For this playground follow Mathias's guidance and solve the problem using your functional language of choice. Contribute your solution by adding a folder named {your twitter handle}+{your language} such as: /bryan_hunter+erlang/

You can push your changes, but it's more fun to submit your solution via pull-request. Time to play!

The data sets

The full data sets train.csv (73.22 mb) and test.csv (48.75 mb) can be found here:
https://www.kaggle.com/c/digit-recognizer/data

For your convenience the smaller data sets (used in the dojo) are contained here in the repo:
training-sample.csv (5000 rows, 9.1 mb)
test-sample.csv (500 rows, 919 kb)