Stabilité et instabilité du risque minimax pour des variables indépendantes équidistribuées
\(X_ 1,...,X_ n\) are independent and identically distributed random variables, with probability law \(P_{\theta}\), \(\theta\) being an unknown parameter. The possible values of \(\theta\) form a given set S. The minimax risk \(R_ n(S)\) is defined as \(\inf_{T}\sup_{\theta \in S}E_{\theta}(nh^ 2(P_ T,P_{\theta}))\) where T is an arbitrary estimator of \(\theta\) and h is the Hellinger distance: \(h^ 2(P,Q)=1- \int \sqrt{dPdQ}.\) Suppose \(S^*_ n\) is a perturbation of S such that any point of \(S^*_ n\) is within \(\epsilon_ n\) of some point of S. This paper investigates how much \(R_ n(S^*_ n)\) will differ from \(R_ n(S)\). It is shown that if \(\epsilon_ n=cn^{-{1\over2}}\), there exists a constant K(c) independent of n and S such that \(R_ n(S^*_ n)<K(c)R_ n(S)\).
- Robust Bayes-like estimation: rho-Bayes estimation
- Learning from MOM's principles: Le Cam's approach
- Robust sub-Gaussian estimation of a mean vector in nearly linear time
- Robust statistical learning with Lipschitz and convex loss functions
- Robust machine learning by median-of-means: theory and practice
- Statistical estimation with model selection
- scientific article; zbMATH DE number 4024495 (Why is no real title available?)
- Estimation of the transition density of a Markov chain
- scientific article; zbMATH DE number 4121209 (Why is no real title available?)
- A new method for estimation and model selection: \(\rho\)-estimation
- Robust high dimensional learning for Lipschitz and convex losses
- scientific article; zbMATH DE number 3892301 (Why is no real title available?)
- Estimator selection with respect to Hellinger-type risks
- Robust subgaussian estimation with VC-dimension
This page was built for publication: Stabilité et instabilité du risque minimax pour des variables indépendantes équidistribuées
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q795430)
