ProbXiv
Explorersign in
Problem archiveProblem record

Statement

Let HH be a dd-regular multigraph on kk vertices, and let ε>0\varepsilon>0. Write Lm(H)\mathcal{L}_{m}(H) for an mm-sheeted lift of HH. The operators Sεrand\mathcal{S}_{\varepsilon}^{\mathrm{rand}}, S~εrand\widetilde{\mathcal{S}}_{\varepsilon}^{\mathrm{rand}}, Sεadv\mathcal{S}_{\varepsilon}^{\mathrm{adv}}, and S~εadv\widetilde{\mathcal{S}}_{\varepsilon}^{\mathrm{adv}} denote, respectively, random, respectful random, adversarial, and respectful adversarial noise as defined in the source. Let G(n,d)\mathcal{G}(n,d) denote a uniformly random dd-regular graph on nn vertices, and let G((n2,n2),d)\mathcal{G}((\frac n2,\frac n2),d) denote the corresponding random bipartite dd-regular graph. Hardness of detecting noisy lifts. If HH is Ramanujan, then there is no polynomial-time algorithm that achieves strong detection between SεrandLm(H)\mathcal{S}_{\varepsilon}^{\mathrm{rand}}\mathcal{L}_{m}(H) and G(n,d)\mathcal{G}(n,d). The same holds when Sεrand\mathcal{S}_{\varepsilon}^{\mathrm{rand}} is replaced by any of S~εrand\widetilde{\mathcal{S}}_{\varepsilon}^{\mathrm{rand}}, Sεadv\mathcal{S}_{\varepsilon}^{\mathrm{adv}}, or S~εadv\widetilde{\mathcal{S}}_{\varepsilon}^{\mathrm{adv}}. If HH is bipartite Ramanujan, then there is no polynomial-time algorithm that achieves strong detection between Sεrand,biLm(H)\mathcal{S}_{\varepsilon}^{\mathrm{rand},\mathrm{bi}}\mathcal{L}_{m}(H) and G((n2,n2),d)\mathcal{G}((\frac n2,\frac n2),d). The same holds when Sεrand,bi\mathcal{S}_{\varepsilon}^{\mathrm{rand},\mathrm{bi}} is replaced by any of S~εrand,bi\widetilde{\mathcal{S}}_{\varepsilon}^{\mathrm{rand},\mathrm{bi}}, Sεadv\mathcal{S}_{\varepsilon}^{\mathrm{adv}}, or S~εadv\widetilde{\mathcal{S}}_{\varepsilon}^{\mathrm{adv}}. These conjectures are intended to imply hardness for certifying quantities that are lift-monotone, such as the independence number divided by the number of vertices, because such quantities can increase under graph lifts.

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.