Interactive · How machines learn

How Machines Learn

No rules are written by hand. The model sees labeled examples, adjusts its internal weights to reduce error, then predicts a label for something it has never seen — with a confidence, not a certainty.

1 · Labeled Data

examples + correct answers

2 · Training

adjust weights to reduce error
epoch 0 / 24
error (loss) ↓

3 · Prediction

unseen input → probabilities
?