Gradient Descent & Optimizer Trajectory Simulator

Loss Surface

Preset Surface

💡 Interaction: Click anywhere on the 2D Contour Map to set a start vector (x,y) and spawn optimizer paths simultaneously.

Hyperparameters

Learning Rate (α) 0.050

Momentum / β1 0.90

RMSprop Decay / β2 0.99

Weight Decay (L2) 0.010

Max Iterations 100

Active Optimizers

SGD Momentum RMSprop Adam

Loss Surface & Trajectories Start: [-1.50, 1.00]

Convergence Dynamics (Loss vs Step) Step: 0/100

SGD 0.0000 Pos: [0.00, 0.00]

Momentum 0.0000 Pos: [0.00, 0.00]

RMSprop 0.0000 Pos: [0.00, 0.00]

Adam 0.0000 Pos: [0.00, 0.00]

Step-by-Step Gradient Vector Inspector

Optimizer	Current Position (x, y)	Raw Gradient ∇f	Velocity / Accum (v)	Effective Step Δ

PROOF STATUS: READY

DEFAULT SURFACE: saddle_point

START VECTOR: [-1.5, 1.0]

CURRENT STEP: 0

FINAL LOSS (ADAM): N/A