Perceptron — 2D Decision Boundary

Class A (+1)

Class B (−1)

boundary w·x+b=0

Data

Click mode

Training

Learning rate: 0.10

Status

Epoch0

Misclassified0

w₁0.000

w₂0.000

bias b0.000

Convergedno

Perceptron learning rule: for point x with target t (+1/−1), predicted p = sign(w·x + b). If p ≠ t, update w ← w + lr·(t − p)·x, b ← b + lr·(t − p). Repeated over the whole dataset (an epoch), this converges to a separating line whenever the data is linearly separable.