Neural Networks
Every modern AI model — from image recognizers to ChatGPT — is built on the same core idea: layers of simple, connected units called neurons.
A simple neural network
Neuron
A tiny unit that takes inputs, weighs them, and passes a signal forward.
Weight
A number that controls how much influence one neuron has on the next — this is what 'training' adjusts.
Activation Function
Adds non-linearity so the network can learn complex patterns, not just straight lines.
Backpropagation
The algorithm that adjusts every weight after each mistake, so the network gradually improves.
How training actually works
A network starts with random weights and makes terrible predictions. Each time it sees a labelled example, it measures how wrong it was (the 'loss'), then nudges every weight slightly in the direction that would have reduced that error — a process called backpropagation. Repeated across millions of examples, this slow nudging is what makes the network 'learn'.
Key takeaways
- Neural networks are layers of simple neurons connected by weights.
- Training means adjusting weights via backpropagation to reduce prediction errors over many examples.
- Deep Learning simply means using networks with many hidden layers, letting the model learn increasingly abstract features.
Check your understanding
0/2 answered1.What does 'training' a neural network actually adjust?
2.'Deep Learning' refers to neural networks with many hidden layers.
Lesson summary
Neural networks learn by repeatedly adjusting the weights between simple, layered neurons via backpropagation — 'deep' just means many layers.
AI-generated notes