03
5 minute interactive lab

Activation functions

Why introduce a bend into the network?

YOU’LL LEAVE KNOWINGNonlinearity lets a network model more than a straight line.
THE EXPERIMENT

Bend the signal without breaking it.

Move the controls. Watch the internal state respond. The explanation follows the behavior—not the other way around.

Runs the real small-scale equationThe displayed values are calculated from the formula shown in this lesson.
ACTIVATIONChoose the decision curve

Negative evidence becomes zero; positive evidence passes through.

relu(-1.8)0.00
THE MENTAL MODEL

Nonlinearity lets a network model more than a straight line.

Without an activation function, stacking layers still collapses into one linear transformation. ReLU, sigmoid, and tanh each introduce a useful bend with different output behavior.

IN DEVELOPER TERMSactivated = f(weighted_signal)
GO DEEPER

Check the model against the source.

Connectionism optimizes for intuition, then points you to the rigorous treatment.

  1. Stanford CS231nActivation functions
  2. Dive into Deep LearningMultilayer Perceptrons
NEXT CONNECTION

How does an input become an output?

The forward pass

Continue to lab 04