How a neural network learns
In short: Predict, measure the error, work out each weight's share of it, adjust the weights, repeat.
1. The network is shown a training example. 2. Layers of connected artificial neurons turn the example into a prediction. 3. A loss function turns the gap to the right answer into one number: the error. 4. Backpropagation works out how much each weight contributed to the error. 5. Gradient descent nudges every weight a step in the direction that shrinks the error.
Sources: Wikipedia, “Artificial neural network (Wikipedia)” (CC BY-SA 4.0); Wikipedia, “Backpropagation (Wikipedia)” (CC BY-SA 4.0); Wikipedia, “Gradient descent (Wikipedia)” (CC BY-SA 4.0); Wikipedia, “Loss function (Wikipedia)” (CC BY-SA 4.0). A simplified picture of training.
Opens the Intellect+ feed with this card first. Swipe up for the next one.
Sources
- Source: Wikipedia: Artificial neural network (Wikipedia) · CC BY-SA 4.0
- Also: Backpropagation
- Also: Gradient descent
- Also: Loss function
To cite this card, link to https://intellectplus.co.uk/app/how-a-neural-network-learns.
How we source, check and correct cards: Editorial standards.
General information, not professional advice. Spotted an error? Flag the card or see corrections and complaints.
Intellect Plus is a visual educational discovery platform for exploring fascinating facts and ideas. Brain Food, Not Brain Rot. What is Intellect Plus?
