kyrie20666 / dropout_dropconnect Goto Github PK

This folder contains the following files and directories:

1) Results: results of the experiments performed
			
	a) csv output files - mnist. The csv output files contain the output predictions of the test data and the training and testing cost per epoch, testing error.
			
	b) ipython notebook files
			
	c) figures used in the report

2) convo_2_layers.py - 2 layer CNN and 1 mlp layer

3) convo_3_layers.py - 3 layer CNN and 1 mlp layer

4) mlp_dropout_dropconnect.py - classes for implementation of nodrop, dropout, dropconnect

5) nn.py - implementation of individual layers in the neural network

6) utils.py - functions for data augmentation and loading the datasets