Adaline vs Perceptron

See why continuous loss functions and gradient descent changed everything.

Parameters

Learning rate (η): 0.5

Standardize features

Actions

Epoch: 0

Perceptron 1957

Errors: —

Accuracy: —

Adaline 1960

MSE: —

Accuracy: —

Loss Landscape

Perceptron

Adaline

Darker = lower loss. Watch Adaline follow the gradient smoothly.

Convergence

Perceptron errors

Adaline MSE

Perceptron jumps discretely. Adaline descends smoothly.

Ready to train