Loss Landscape & Optimizer Dynamics p5.js Bench

Rosenbrock Ravine
f(x,y) = (1-x)² + 100(y-x²)²
Click canvas to set release point
Bench State: Adam smoothly navigates the narrow curved Rosenbrock ravine in under 120 steps while standard SGD stutters on the ill-conditioned gradient scale.
Loss Surface Topography
Global Hyperparameters
Learning Rate (α) 0.0100
Momentum Beta (β₁) 0.90
RMSprop/Adam Beta2 (β₂) 0.999
Active Optimizers
Live Telemetry
Opt Steps Loss f(x,y) ||∇f||