3D Gradient Descent Explorer

3D Surface — drag to rotate, scroll to zoom

Contour Map — click to set start

Loss Function

Optimizer Algorithm

Learning Rate (α)0.050

Start X-3.0

Start Y3.0

Speed5

Step: 0 Algorithm: Batch GD
Position: (−3.0, 3.0)
Loss: — |∇|: —
Status: Ready

Start Path Current

Batch GD: exact gradient SGD: noisy (1 sample) Mini-batch: moderate noise

Quadratic Bowl: f(x,y) = x²+y². Convex, single minimum at (0,0).