Artificial neuron, explained simply
Neural networks power many modern AI tools. They are built from lots of tiny units called artificial neurons. Each one does a simple sum.
A neuron takes some input numbers and multiplies each by a weight, which says how important that input is. It adds them up, adds one more number called the bias, and then an activation step decides the output, for example "yes" if the total is above zero.
Learning means adjusting the weights and bias. When the neuron gets an example wrong, each weight is nudged a little in the direction that would have made the answer better. Repeat that over many examples and the neuron improves. Big networks do the same with millions of neurons.
Example: Should Pip go outside?
- Inputs: sunny = 1, raining = 0.
- Weights: sunny × 2, raining × −3. Bias: −1.
- Sum: (1 × 2) + (0 × −3) − 1 = 1. Above zero, so the output is "yes, go outside".
- If it rains: (1 × 2) + (1 × −3) − 1 = −2. Below zero, so "stay in".
Try this
- Build a "should I have ice cream?" neuron with two inputs. Choose weights you think are sensible and test a few days.
- Change one weight and see which decisions flip.
- Pretend to be the neuron: when you get an example wrong, nudge a weight and try again.