The output layer

The output layer is basically the output value of the network and is formed depending on the problem setting. In unsupervised learning, such as encoding or decoding, the output can be the same as the input. For classification problems, the output layer can have n neurons for n-way classification and utilize a softmax function to output the probability of being each class. Overall, the output layer maps to your target space and the perceptron in it would change accordingly, based on your problem setting.