3 · A peek at where this is heading
Two lines of Python you'll write over and over in ML — and the picture they paint.
Every supervised ML pipeline starts by splitting data into a train set and a test set.
Without this:
Without a held-out test set, you can't tell whether your model learned anything or just memorized.
Take a deep breath. We're three lessons in and you're about to see the single most important pattern in supervised ML: the train/test split. The scene below shows what happens when you call one line of scikit-learn.
In Python that's literally two lines:
Python (in browser)
80/20 split with a fixed random seed so the result is reproducible.
Python runs entirely in your browser via Pyodide (~6 MB on first Run, cached after).
Why split? It's about generalization — the math is gradient descent on unseen data.
Don't worry if scikit-learn looks intimidating right now. By the end of Track 5 (Classical Supervised ML) you'll write this in your sleep.
- Train/test split is the bedrock of every honest ML experiment.
- A fixed random seed makes results reproducible — a habit you should keep.
Literally every supervised learning notebook on the planet starts with `train_test_split(...)`.
If you remove it: You can't trust a model evaluated on the data it trained on — the metric tells you nothing about real-world performance.