Home▸Mathematics▸Search Space Analysis: Hyperparameter Sampling Strategies (2D)

Search Space Analysis: Hyperparameter Sampling Strategies (2D)

2D companion to the 3D hyperparameter search-space lab: grid, random, Latin-hypercube and Bayesian samplers explore the identical loss landscape as a top-down heatmap with a live convergence chart.

Mathematics2DAdvanced60 FPS📱 Mobile-adapted⇄ 3D version

This 2D companion drives the identical loss-landscape function and sampling strategies as the 3D version — grid, random, Latin hypercube and Bayesian-guided search over two hyperparameters — through a flat top-down heatmap built for reading the math directly: color encodes loss at every point, sample markers appear as each configuration is evaluated, a glowing marker tracks the best point found so far, and a live convergence chart plots best-loss-so-far against samples evaluated so you can directly compare how efficiently each strategy explores the space.

⚙ Under the hood

2D companion to the 3D hyperparameter search-space lab: the identical closed-form loss landscape and grid/random/Latin-hypercube/Bayesian-guided samplers rendered as a top-down heatmap with sample markers, a best-point tracker, and a live best-loss-vs-samples convergence chart instead of a rotating 3D mesh.

hyperparametersoptimizationsearch spacebayesian optimizationlatin hypercubemathematics2d-simulation

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