There are several types of layers that our neural network can create. Since for this example we are dealing with a CNN, the first layer must be a fully convolutional layer. The first layer to be added must be an input layer or an exception will be thrown. After the input layer is added, the next layer must be a classification layer. This is a fully connected layer with a defined neuronal count, bias, bias gradient, bias preference values, and L1 and L2 decay multipliers. If the layer is a ReLU layer, ReLU nonlinearity element wise will be implemented such that x -> max(0,x), with the output being zero to infinity.