Learning lesson 01 · Learn from a mistake
01 · Smallest useful mechanism
Give one artificial neuron numerical inputs, multiply them by weights, add a bias, and threshold the score. The output is a prediction. During training, a labelled example reveals whether that prediction was wrong. The perceptron update adds a small, directed correction to the weights, so the same mistake becomes less likely on the next pass.
A perceptron stores a linear decision boundary in its weights and bias; an update nudges that boundary only when an example is misclassified.
02 · Experiment
Exact perceptron updates · deterministic teaching data
Every input, score, mistake, and weight update is declared in the browser. The values demonstrate the learning rule rather than reconstructing a production model.
03 · Reason from evidence
Your brief: Write a short diagnosis that separates a bad update from stale parameters and an impossible boundary. Do not suggest more training until you know which failure you have.
Evidence your answer should use
There is no reveal button here. A strong response names the claim, the evidence that would test it, and what the evidence cannot establish.
Your learning artifact
Stored only in this browser. No account required; course reset does not delete it.
04 · Check your understanding
Next: The update is compact because both the example and the model are vectors. Next we make that geometry explicit.
Decision space
Next training example
false OR false
Target: −1
Current parameters
Ready for the first example.
Press Step to reveal the score, prediction, and parameter update.
