Hands-On Neural Networks
上QQ阅读APP看书,第一时间看更新

Activation functions

So far, you have seen two different activation functions: a step function and a sigmoid. But there are many others that, depending on the task, can be more or less useful.

Activation functions are usually used to introduce non-linearity. Without it, we will only have a linear combination of input going through another linear function.

We will now look at a few activation functions and their code in Keras, in detail.