Testing classification of MNIST digits in Ruby.
Includes a Sinatra app that uses a trained ruby-fann neural network to predict digits drawn on a element. The neural network was trained on all 60,000 training examples with 1 hidden layer of 300 neurons, and successfully classified ~93% of the test set.
MNIST data files not included in this repo.