The sigmoid function is frequently used as the activation function in the output layer of binary classification models. Because its output is bounded within the range $$(0, 1)$$, the result can be directly interpreted as a probability, which can then be easily thresholded and converted into discrete class predictions.

University of Michigan - Ann Arbor

Google

- The hyperbolic tangent (tanh) function outputs values in the range $$(-1, 1)$$, while the sigmoid function outputs values in the range $$(0, 1)$$.
- tanh generally works better than sigmoid as a hidden-layer activation function, because its output is centered around 0 rather than 0.5, which tends to make training/optimization faster.
- Sigmoid is still preferred for the output layer of binary classification models, since its $$(0, 1)$$ range maps directly to a class probability.

Learn Before

Related