Sigmoid Activation Function - Deep Learning Dictionary
text
Sigmoid Activation Function - Deep Learning Dictionary
In a neural network, an activation function applies a nonlinear transformation to the output of a layer.
One activation function, called \(\text{sigmoid}\), maps its supplied inputs to a value in the interval \((0,1)\). For a given value \(x\) passed to \(\text{sigmoid}\), we define
$$\text{sigmoid}(x)=\frac{e^{x}}{e^{x}+1}$$
The table below summarizes how sigmoid
transforms its input.
Input | Sigmoid Output |
---|---|
Most negative values | A value very close to \(0\) |
Most positive values | A value very close to \(1\) |
Values relatively close to \(0\) | A value between \(0\) and \(1\) |
When sigmoid is used as an activation function following a layer in a neural network, it accepts the weighted sum of outputs from the previous layer and transforms this sum to a value between \(0\) and \(1\).
Intuitively, we can think of a given node's activated output as being "more activated" the closer it is to the upper limit \(1\), and similarly "less activated" the closer it is to \(0\).
We can therefore think of \(\text{sigmoid}\)'s output as being a probability of activation for a given node.
quiz
resources
updates
Committed by on