← All problems

Particle-swarm stagnation law is non-Gaussian

19 years old · traced to

Can the limiting position distribution of the canonical one-dimensional particle-swarm optimizer during stagnation be Gaussian?

References

Progress summary

Refreshed
Open · solution reported

A 2026 manuscript claims that the stagnation distribution cannot be Gaussian, but the result has not yet been independently confirmed.

The problem asks whether the limiting position distribution of a one-dimensional particle-swarm optimizer during stagnation can be Gaussian. It appears as Problem 18, “Bell curve,” on the 2026 Particle Swarm Central problem list.

2026 manuscript: claimed negative resolution

The manuscript claims that, throughout the stated nondegenerate mean-square stability region −1<w<1-1<w<1, c>0c>0, and 12(1−w2)−c(7−5w)>012(1-w^{2})-c(7-5w)>0, no invariant or limiting position marginal is Gaussian. Its argument compares stationary moments through order eight: fourth- and sixth-order tests admit a candidate, but the eighth-order condition excludes it; the reported standardized eighth moment is 106.579793218089576…106.579793218089576\ldots, rather than the Gaussian value 105105. It further claims that the joint stationary law (Xt+1,Xt)(X_{t+1},X_t) is always non-Gaussian when the noise variance satisfies σϕ2>0\sigma_{\phi}^{2}>0.

Current status (as of 2026): A preprint claims the problem is solved negatively, but independent verification is still absent.

Sources

Solutions 0

No solutions have been posted yet.