WiSARD-MNIST

(ASSIGNMENT) Use WiSARD neural network to identify handwritten characters from MNIST database.

requirements

• Python 2.7
• PyWANN library: https://github.com/firmino/PyWANN
• MNIST datasets are included (Resources folder).

how to run

Classifier.py [train_image_file] [train_label_file] [test-image-file] [test-label-file] [options]

options:

• --num_bits_addr INT: sets number of address bits (vanilla WiSARD parameter)
• --bleaching BOOL: turns bleaching on and off (default: True)
• --bleaching_initial_value INT: initial value for bleaching threshold (default: 1)
• --bleaching_confidence FLOAT: confidence value used in bleaching ties (default: 0.1)

info

• Made by: Silvia Pimpão, Vinícius Garcia
• For: Redes Neurais Sem Peso (2017/2)