MACHINE LEARNING / TECHNOLOGY

Learning Plane: Gradient Descent

Train a logistic classifier and see its decision boundary evolve with each update.

LIVE SIMULATIONSIMULATING
Your browser needs HTML5 Canvas.
MODEL ENGINE / 12REAL-TIME MODEL
SIM TIME 0.0s
Cross entropy—
Accuracy—
Training steps—

✦ The science behind the simulation

Logistic regression uses p = 1/(1+e^(−w·x)) and minimises cross-entropy. The weights are updated using gradients of the prediction error over the training samples.

Under the hood

Logistic sigmoid · cross-entropy · batch gradient descent

Made with 🤖 ChatGPT