MNIST-digit-classifier

MNIST digit classifier using my own neural network library. version 1.0
find the library here.

Beta version 1.0

the network can currently classify the MNIST data with approximately 95% percent accuracy. It is tested for a variable number of times. currently the visualization only uses the command line to display the error and percent of training complete, then, in the testing loop it prints the correct answer, the guess and whether or not it is correct.

To do

• add visualizations of the digits and interactivity for user drawings.
• improve the networks initial hyperparameters for better accuracy.
• save a very accurate model for later use.
• improve the neural network library