variance reduction
Stabilizing noisy estimates
Variance reduction refers to statistical techniques that decrease the randomness in estimates, making simulations or measurements more precise without requiring more samples.
Common in Monte Carlo methods and experimental design, these techniques exploit known relationships in data to produce tighter, more reliable results: Importance sampling is a variance reduction technique that focuses computational effort where it matters most.
- antitheticDirectly opposed in nature
- jackknifeFolding pocket blade
- shrinkageDecrease in size
- quality controlMaintaining production standards
- bootstrappingSelf-starting without external input
Six Sigma · cross-validation · stress testing · standard error · error bars · law of large numbers · replication · importance · efficiency · stratification · noise filtering · hedging · sampling · sample size · regularization · contingency planning · noise · uncertainty
- Monte Carlorandom sampling
- renderingray tracing, shaders, shader, reflections, shadows
- tree ensemblesdecision tree, decision trees, ensemble, bagging, boosting
- samplingsample collection, random sample, random sampling, representative sample, subsampling
- Transport PhysicsBoltzmann equation, diffusion, scattering, flux, cross sections