ProbXiv
Explorersign in
Problem archiveProblem record

Statement

Let WW be the recurrent operator, let ρ(W)\rho(W) denote its spectral radius, and let R^\hat{R} be a non-trivial unitary input encoding. For an input sequence {uτ}\{\mathbf{u}_{\tau}\} and input encoding RR, write

ρeff(W;R,{uτ})≡Eτg[ρ(WR(uτ))]\rho_{\mathrm{eff}}(W;R,\{\mathbf{u}_{\tau}\})\equiv \mathbb{E}^{g}_{\tau}[\rho(WR(\mathbf{u}_{\tau}))]

for the effective input-driven spectral radius, where Eg\mathbb{E}^{g} denotes a geometric mean. The nonstationary echo state property (ESP) is expressed by Eq.

. **Existence of unitary input encoding for the nonstationary ESP.** If $\rho(W)<1$, then there always exists a non-trivial unitary input encoding $\hat{R}$ such that Eq.

is satisfied. Furthermore, the convergence rate with respect to tt is

O(ρeff(W;R,{uτ})t).O\left(\rho_{\mathrm{eff}}(W;R,\{\mathbf{u}_{\tau}\})^t\right).

This conjecture proposes that a recurrent operator with spectral radius less than one admits a unitary input encoding ensuring the nonstationary ESP, with a rate governed by the effective input-driven spectral radius. The supplied text does not establish the claim or indicate that it has been resolved.

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.