Empirical Likelihood
A nonparametric statistical inference method that constructs confidence intervals and hypothesis tests by maximizing the empirical likelihood function, without assuming a specific parametric form for the data distribution. It is used in bandit algorithms to construct finite-time and tail risk guarantees.
Timeline (last 90 days)
Estimation in mode regression is usually implemented via kernel density estimation or empirical likelihood methods, and its statistical properties are more complex than those of mean regression, with the convergence rate depending on the curvature of the density near the peak.
The EL-F algorithm extends g-Forcing through Empirical Likelihood, and while preserving asymptotic optimality, gives a finite-time upper bound on expected regret and a characterization of tail risk.
All Facts (2)
Estimation in mode regression is usually implemented via kernel density estimation or empirical likelihood methods, and its statistical properties are more complex than those of mean regression, with the convergence rate depending on the curvature of the density near the peak.
50%UnverifiedThe EL-F algorithm extends g-Forcing through Empirical Likelihood, and while preserving asymptotic optimality, gives a finite-time upper bound on expected regret and a characterization of tail risk.
50%