The building block
A perceptron is the simplest neuron. It multiplies each input by a weight, adds them up with a bias, and passes the sum through an activation function. That whole sequence turns several numbers into one output.
- Weights decide how much each input matters.
- Bias shifts the threshold for firing.
- Activation bends the weighted sum into a useful range.
Why activation matters
Without an activation function a stack of neurons collapses into a single linear map, no matter how many layers you add. The nonlinearity is what lets networks model curved boundaries and complex patterns.
- Sigmoid squashes values into zero to one.
- Tanh centers output around zero.
- ReLU passes positives through and zeros out negatives.
Reading the flow
The neuron is a tiny pipeline: combine inputs linearly, then apply a nonlinear gate. Layering many of these gates is what gives deep networks their expressive power.
Key idea
A perceptron forms a weighted sum of inputs plus a bias and passes it through an activation function, and that nonlinearity is what lets stacked neurons model patterns a single line cannot.