Interactive · How machines learn

How a Model Is Trained

The same pipeline runs behind almost every AI system. Data comes in, gets cleaned and split, the model trains by finding the statistical patterns that link inputs to answers, and then it is graded on examples it has never seen. If it fails, the loop starts again with better data. Tap any stage to see what it does, or press Run the pipeline.
01
{{ICON_DB}}
Collect
Raw labeled examples arrive
02
{{ICON_CLEAN}}
Clean
Fix errors, drop junk
03
{{ICON_SPLIT}}
Prepare
Split: train vs. hidden test
04
{{ICON_TRAIN}}
Train
Learn the patterns in the data
05
{{ICON_EVAL}}
Evaluate
Grade it on unseen examples
06
{{ICON_SHIP}}
Model ready
Predict on new inputs
not good enough? → back for more & better data
ReadyPress Run the pipeline to watch a dataset travel from raw rows to a trained model — or tap any coloured stage above to see what it does.
Tip: hover a stage to lift it, click to read it.