📈 Hyperparameter Search Compared: Grid vs Random vs Bayesian Optimization (3D)
Walk over a 3D validation-loss landscape while grid, random and Bayesian search place their trials: regret, expected improvement and a Gaussian-process surface with live numbers.
⚙ Under the hood
Walk over a 3D validation-loss landscape while grid, random and Bayesian optimisation place their trials. Simple regret, a Gaussian-process surface, expected improvement and a 40-landscape average run live, with seven independent numerical checks and a Blender-built lab model.
hyperparameter tuninggrid searchrandom searchBayesian optimizationGaussian processexpected improvementregretmachine learning
3D · Three.js / WebGL renderer · 60 FPS target · runs fully client-side, no install