Neural networks: how the machine learns

A neural network is inspired by the brain: millions of small "neurons" in layers. Each neuron is a simple formula: input × weight. Training means tuning billions of weights so the error shrinks. Show a picture, the net says "dog", correct is "cat" — weights nudge toward "cat". Repeat a million times — and the net "knows".

• ●Input layer: pixels, words, numbers
• ●Hidden layers: find features — edges, shapes, meanings
• ●Output: the answer — "cat", translation, next move