Home▸AI & Machine Learning▸Gradient Descent 2D: Contour-Map Optimiser Race

Gradient Descent 2D: Contour-Map Optimiser Race

Watch SGD, Momentum, RMSprop and Adam race across a top-down contour map of classic loss-landscape test functions. Adjust learning rate, momentum and gradient noise and compare convergence paths live.

AI & Machine Learning2DModerate60 FPS📱 Mobile-adapted⇄ 3D version

This 2D companion drives the same four optimiser update rules as the 3D version — SGD, Momentum, RMSprop and Adam — but plots them on a flat, top-down contour heatmap instead of an orbiting 3D mesh, which makes it easier to compare path shapes and convergence speed across the Rosenbrock, saddle, Beale and Himmelblau test functions at a glance.

⚙ Under the hood

Analytic gradients for four classic optimisation benchmarks, descended by SGD, Momentum, RMSprop and Adam and rendered as a live contour-map race with trails, a live readout panel and an adjustable learning rate, momentum and gradient-noise.

gradient descentoptimizationmachine learningsgdadamcontour map

2D · HTML5 Canvas 2D · 60 FPS target · runs fully client-side, no install