Choosing activation functions for feedforward neural networks

For simplicity, we have only discussed the sigmoid activation function in context of multilayer feedforward neural networks so far; we used in the hidden layer as well as the output layer in the multilayer perceptron implementation in Chapter 12, Training Artificial Neural Networks for Image Recognition. Although we referred to this activation function as sigmoid function—as it is commonly called in literature—the more precise definition would be logistic function or negative log-likelihood function. In the following subsections, you will learn more about alternative sigmoidal functions that are useful for implementing multilayer neural networks.

Technically, we could use any function as activation function in multilayer neural networks as long as it is differentiable. We could even use linear activation functions such as in Adaline (Chapter 2, Training Machine Learning Algorithms for Classification). However, in...