6. Connecting Back to Teachable Machine¶
Every idea from this unit is already sitting inside Teachable Machine, under its Advanced settings panel, visible right before you click Train Model.
| Teachable Machine setting | What you now understand about it |
|---|---|
| Epochs | How many full passes through your training images (Lesson 2) |
| Batch Size | How many examples it looks at before each weight update |
| Learning Rate | Exactly Lesson 3 — too high and training destabilizes, too low and it crawls |
Teachable Machine hides the hidden-layer structure from you entirely, but it's there — a much larger network than anything you built in Playground, doing the exact same weighted-sum-and-adjust process on image data instead of two simple numbers.
Why this unit mattered¶
Unit 2 let you use a neural network. This unit let you see inside one — what overfitting actually looks like, why a learning rate matters, why some problems need more network than others. That understanding is exactly what makes Unit 2's symptoms (a model that memorizes instead of generalizing, one that needs more varied data) make sense mechanically instead of just by rule of thumb.
Try it
Open a Teachable Machine project from Unit 2, open its Advanced settings before training, and identify which Playground concept each one corresponds to using the table above.
Tip
You don't need to do the math by hand to use any of this well — understanding what each setting does, even without the formulas behind it, is enough to train better models and debug them when they don't work.