ProbXiv
Explorersign in
Problem archiveProblem record

Statement

a) is there any other estimator enjoying this univsersal estimation property? b) is this property satisfied by the empirical mean is due to the fact that the empirical mean is a sufficient statistics (and that it is somehow better than any other estimator)? c) is it possible to extend this result beyond the Gaussian case? is it possible to extend this property even when some data are not all Gaussian and may have been corrupted even in the worse corruption model which is the ǫ-adversarial corruption model? (and up to which proportion ǫ of corruption?) It is the aim of this work to answer these questions. Theorem 2. We assume that a fraction ǫ of the N Gaussian vectors G1 , . . . , GN has been corrupted by an adversary. We denote by X1 , . . . , XN the resulting adervsarially corrupted dataset given to the Tukey median µ̂ defined as N ! 1 X µ̂ \in argminµ\inRd D(µ) where D(µ) = sup I Xi − µ, v \geq 0 . v\inRd N i=1

Record

Comments

Nothing has been published against this problem yet, and nothing has been said about it.

Sign in with an institutional address to take part in the discussion. Reading every thread stays open to everyone.

Sign in

Solve with an agent

Open the statement in a chat, with the problem and the ground rules already written into the prompt.

This opens a third-party site. Nothing is posted back to ProbXiv and nothing you write there is recorded here — what a model gives you is an attempt, which a person still has to check.