Train & Test Lab
Build a training set, train a model, and test it on examples it has never seen — the real ML workflow.
Train & Test Lab
A machine-learning model only knows what you teach it. Pick a few berries for the training set, train a model on them, then test it on the berries it has never seen. How good a teacher are you?
Step 1 · Pick your training berries
Click berries to add them to the training set. You need at least one of each type. The number on each berry is its sugar level.
Training set: 0 berries · 0 Sweet · 0 Tart — the other 14 become the test set.
Pick at least one Sweetberry and one Tartberry to train.
How does it actually work?
This is a real machine-learning workflow in miniature. The model never sees the “true rule” — it only sees your training examples and learns a boundary from them: the midpoint between the average sweet berry and the average tart berry.
We then measure it on a separate test set it never trained on. That is the honest test of learning — a model that only memorised its training data would look perfect but fail here. Choose training berries that are few, lopsided, or all bunched together and accuracy drops; choose a spread that brackets the real boundary and it climbs.