Chapter 3 · Neural networks
A neuron: weights and bias
A neuron multiplies each input by a weight, adds a bias, and passes the sum through an activation.
Chapter 3 · Neural networks
A neuron multiplies each input by a weight, adds a bias, and passes the sum through an activation.
1. Inputs arrive (1 = yes, 0 = no)
Weights decide how much each input matters. The bias is the neuron's starting lean. The wire thickness shows the weight: green wires push towards yes, coral wires push towards no.
An input is just a number. Here: dark clouds? forecast says rain? a friend carries an umbrella? Each is 1 for yes and 0 for no.
Tap the inputs to switch them, then move the weights and bias. Can you make the neuron ignore dark clouds? Can you make it say yes even with no inputs on?
1×1.5 + 0×2 + 1×0.5 + (-1) = 1 → squash → 0.73
☂️ The neuron says: take an umbrella.
A teaching version with three yes/no inputs. The weights here are set by you; in a real network they are learned from data.
A weight of 0 on an input means…
What does a strongly negative bias do?
What is learned in a neuron?