English

Optimizing Monotone Functions Can Be Difficult

Neural and Evolutionary Computing 2015-03-17 v2

Abstract

Extending previous analyses on function classes like linear functions, we analyze how the simple (1+1) evolutionary algorithm optimizes pseudo-Boolean functions that are strictly monotone. Contrary to what one would expect, not all of these functions are easy to optimize. The choice of the constant cc in the mutation probability p(n)=c/np(n) = c/n can make a decisive difference. We show that if c<1c < 1, then the (1+1) evolutionary algorithm finds the optimum of every such function in Θ(nlogn)\Theta(n \log n) iterations. For c=1c=1, we can still prove an upper bound of O(n3/2)O(n^{3/2}). However, for c>33c > 33, we present a strictly monotone function such that the (1+1) evolutionary algorithm with overwhelming probability does not find the optimum within 2Ω(n)2^{\Omega(n)} iterations. This is the first time that we observe that a constant factor change of the mutation probability changes the run-time by more than constant factors.

Keywords

Cite

@article{arxiv.1010.1429,
  title  = {Optimizing Monotone Functions Can Be Difficult},
  author = {Benjamin Doerr and Thomas Jansen and Dirk Sudholt and Carola Winzen and Christine Zarges},
  journal= {arXiv preprint arXiv:1010.1429},
  year   = {2015}
}

Comments

Preliminary version appeared at PPSN XI. Compared to version 1, a small bug in the constants was fixed ($\gamma$ is slightly larger now, thus ensuring that $\gamma$ is now strictly larger than $\rho$ in Lemma 5)

R2 v1 2026-06-21T16:25:13.493Z