1. Neurons and Layers¶
TensorFlow Playground is a free, visual simulator that shows you exactly what's inside the kind of model you trained in Unit 2 — no installation, nothing to export, just a webpage.
The three kinds of layer¶
| Layer | What it holds | Where you see it in Playground |
|---|---|---|
| Input layer | The raw features going in | The leftmost column (X1, X2, and a few calculated ones) |
| Hidden layer(s) | Where the actual "thinking" happens | The middle columns — you can add or remove both layers and neurons within them |
| Output layer | The final answer | The colored result on the right, matching the two classes in the data |
A single neuron does something simple: it takes its inputs, multiplies each by a weight, adds them up, and passes the result through a small math function. Nothing about one neuron alone is impressive — the power comes from stacking many of them together.
This is the same kind of thing as Unit 2¶
Your Teachable Machine model from Unit 2 is a neural network too — just with a much bigger input layer (every pixel of an image, not just two numbers) and many more neurons. Everything in this unit is the same idea, deliberately shrunk down small enough to actually watch.
Try it
Open TensorFlow Playground, select the Circle dataset, and leave everything at its defaults. Without training anything yet, point to the input layer, a hidden layer, and the output on screen.
Tip
The colored squares inside each hidden-layer neuron are a preview of what that specific neuron has learned to detect — even before you fully understand the math, you can see that different neurons pick up on different patterns.