5.3 Artificial Neurons and Layers
An artificial neuron multiplies inputs by weights, adds them up, adds a bias, then decides how strongly to "fire".
Chapter 5: Deep Learning and Neural Networks
Zoom into one neuron. It takes several inputs, multiplies each by its weight, adds them all together, and adds a bias (a little nudge). Then it passes the result through an activation function, which decides how much signal to send on.
A popular activation function, ReLU, is super simple: if the number is negative, output 0; otherwise pass it through. That small bit of non-linearity is what lets networks learn complex patterns instead of only straight lines.
Stack thousands of these neurons into layers and you get something that can recognise faces or write sentences.
In one line
output = activation( w₁·x₁ + w₂·x₂ + w₃·x₃ + bias )
Key takeaways
Neuron = weighted sum + bias + activation
Activation functions add non-linearity
ReLU: negative → 0, positive → keep it