Chapter 5

Perceptron Learning Simulator

Draw data points, then train the perceptron step-by-step or watch it converge automatically

The perceptron learning algorithm updates weights when it makes a mistake: w ← w + η(y − ŷ)x. For a false negative (missed positive point), the boundary rotates toward the point. For a false positive, it rotates away. If the data is linearly separable, this process is guaranteed to converge. Try XOR to see what happens when it isn't!

Errors per Epoch

Controls

Draw Points

Click the canvas to place points, or pick a preset:

Presets
Learning Rate (η)
Training
Current Weights
w₁
0.00
w₂
0.00
b
0.00
Epoch
0
Errors
0
Ready — draw points or pick a preset
Update Rule Cases
CaseyŷUpdate
Correct (pos)11None
Correct (neg)00None
False negative10w + ηx
False positive01w − ηx