1
2
3

The Decider

Decide when a neuron passes its signal on.

Step 1 of 3

What you'll learn in this level

  • What an activation function does inside a neuron
  • How step, sigmoid, ReLU and tanh differ
  • Why a network could barely learn anything without them

What a Neuron Passes On

Neurons don't just pass numbers along; they make decisions! An activation function decides whether a neuron 'fires' (sends a signal) or stays quiet.

Different functions work better for different jobs. Some are like simple on/off switches, others are more gradual.

Common Functions

  1. Step

    Like a light switch: ON or OFF. Good for simple yes/no decisions.

  2. Sigmoid

    A gentle ramp from 0 to 1. Good when you need a probability.

  3. ReLU

    If the number is positive, it stays. If it is negative, it becomes zero. Very popular in modern AI.

  4. Tanh

    Like sigmoid, but from -1 to 1. So the neuron can also vote against.

Now test for yourself how these functions change the signal.

Guess first, then experiment

What happens if you remove the activation function from every neuron?