ProbXiv
Explorersign in
Problem archiveProblem record

Statement

Let σt\sigma_t denote the scalar ensemble standard deviation of the parameter θ\boldsymbol{\theta} at iteration tt, and let Cθ^r,t,Hx^r,tC_{\hat{\boldsymbol{\theta}}_{r,t},H\hat{\mathbf{x}}_{r,t}} be the covariance between the resampled parameter ensemble and the corresponding observation ensemble. Let gg be a non-negative function of σt\sigma_t. Resampling convergence conjecture. There exists a non-negative function g(σt)≥0g(\sigma_t)\geq 0 such that

∣Cθ^r,t,Hx^r,t∣≥g(σt)>0,∀ σt≠0,\left|C_{\hat{\boldsymbol{\theta}}_{r,t},H\hat{\mathbf{x}}_{r,t}}\right|\geq g(\sigma_t)>0,\qquad \forall\,\sigma_t\neq 0,

and g(σt)=0g(\sigma_t)=0 only when σt=0\sigma_t=0. This would rule out the false-convergence mechanism in which the parameter and observation ensembles become uncorrelated without collapsing, so that resampling can preserve convergence-driving covariance whenever the ensemble variance is nonzero.

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.