🎲 Probability Theory in Hyperparameter Optimization 3D
Why random search finds a good hyperparameter region with probability 1-(1-q)^n, and how Gaussian-process Bayesian optimisation with expected improvement compares, on a known two-parameter landscape with Monte-Carlo checks.
⚙ Under the hood
Why random search finds a good hyperparameter region with probability 1-(1-q)^n, and how Gaussian-process Bayesian optimisation with expected improvement compares, on a known two-parameter landscape with Monte-Carlo checks.
hyperparameter optimizationprobabilitybayesian optimizationrandom searchmachine learning
3D · Three.js / WebGL renderer · 60 FPS target · runs fully client-side, no install