yo.py

        #first we feed the input image into a convolutional layer to get a feature map
		h = self.cnn_layer(X, layer_i=0, border_mode="full") ; X = h
        #then we acivate it with a nonlinearity to make sure the math doesn't break (turn all neg numbers to 0)
		h = self.relu_layer(X) ; X = h
        #another CNN layer for more, smaller images (more hierarchical features = better prediction)
		h = self.cnn_layer(X, layer_i=2, border_mode="valid") ; X = h
        #another nonlinearity 
		h = self.relu_layer(X) ; X = h
        #pooling to reduce computational cost, extract the most important features
		h = self.maxpooling_layer(X) ; X = h
        #dropout to prevent overfitting, randomnly turn nodes on and off, force the network to learn multiple
        #pathways for data
		h = self.dropout_layer(X, .25) ; X = h
        #reduce dimensions of an input tensor
		h = self.flatten_layer(X) ; X = h
        #to combine all learned parts of inputs 
		h = self.dense_layer(X, layer_i=7) ; X = h
        #nonlinearity
		h = self.relu_layer(X) ; X = h
        #more dropout
		h = self.dropout_layer(X, .5) ; X = h
        #more combination
		h = self.dense_layer(X, layer_i=10) ; X = h
        #sigmoid to output probability values
		h = self.softmax_layer2D(X) ; X = h
        #make the classification based on the probabitlies
		max_i = self.classify(X)
        #return predicted label
		return max_i[0]