How it works
Model P(class=1) with sigmoid: σ(z) = 1/(1+e⁻ᶻ) where z = w₁x + w₂y + b. The decision boundary sits where P = 0.5.
Minimize cross-entropy loss via gradient descent. No closed-form — pure iteration.
blue = class 0, red = class 1
Model P(class=1) with sigmoid: σ(z) = 1/(1+e⁻ᶻ) where z = w₁x + w₂y + b. The decision boundary sits where P = 0.5.
Minimize cross-entropy loss via gradient descent. No closed-form — pure iteration.