suayder / Mnist-Wisard

This is a simple WISARD implementation and testing with MNIST dataset

Mnist-Wisard

This is a simple WISARD Neural Network implementation and testing with MNIST dataset. The article about this repo can be found here

What do i need?

- Python 3.6
- numpy 1.14.5

How can i run it?

After check yours dependencies, you will need to set up your input parameters, for this edit an json file named Network_Settings.json, where it containd the following fields:

"result":"result/result.dat"

• n_inputs_ram -> this is the number of inputs in each WIZARD RAM
• examples_samples -> Number of samples to train you would like to train
• training_dataset -> Path mnist files to train
• testing_dataset -> Path mnist files to test
• result -> file name to save the result

then you just need run Main.py

Optionally: there is a file test.py where you can make the automatic tests with many settins by time, they will create many files of results, this files you can concatenate running concatenate.py and to complite generate a chart with plot.gnu that use gnuplot to generate the graphic with the results.