The $(1 + 1)$-EA in Dynamic Environments
The authors study the (1+1)-EA in dynamic linear environments and prove sharp thresholds for mutation rates in two models.
The paper introduces a new analysis of the (1+1)-EA in dynamic linear environments and provides a sharp threshold for mutation rates.
Before reading this…
To understand this paper, make sure you know these concepts first:
- Evolutionary algorithmsfind papers →
- Dynamic environmentsfind papers →
Abstract
More Like ThisWe study the $(1 + 1)$-EA in dynamic linear environments, where in every generation selection is performed with respect to a freshly sampled linear function with positive weights. We consider the Dynamic Binary Value problem, where each generation uses a uniformly random permutation of $1,2,4,\dots,2^{n-1}$, and a Uniform weight variant, where the weights are drawn independently from $\mathrm{Unif}(0,1)$. Both of them have recently been integrated into the IOHprofiler platform and empirically studied. For both models we prove a sharp threshold in the mutation parameter $χ$ for mutation rate $χ/n$. Below the threshold, the expected optimisation time is $\mathcal{O}(n\log n)$, whereas above it the runtime becomes $2^{Ω(n)}$. For the Dynamic Binary Value problem in the exponential regime, we also quantify at what distance from the optimum the optimisation process stagnates. We show that there is a second threshold: a distance that is efficiently reached, but reaching any smaller distance takes exponential time. This quantifies and proves previous empirical findings.