Name: Gradient Descent Lab
Author: MySimulator

Optimizer settings

Learning rate: 0.08

Use momentum (β = 0.85)

Click anywhere on the surface to drop the marker there and restart. The surface has two bumps and a shallow valley — a small, deliberately non-convex loss landscape.

State

Step0

Current loss—

Statusready

Loss surface (contour) — click to place the starting point