With apparently all research on estimation-of-distribution algorithms (EDAs) concentrated on pseudo-Boolean optimization and permutation problems, we undertake the first steps towards using EDAs for problems in which the decision variables can take more than two values, but which are not permutation problems. To this aim, we propose a natural way to extend the known univariate EDAs to such variables. Different from a naive reduction to the binary case, it avoids additional constraints. Since understanding genetic drift is crucial for an optimal parameter choice, we extend the known quantitative analysis of genetic drift to EDAs for multi-valued variables. Roughly speaking, when the variables take $r$ different values, the time for genetic drift to become significant is $r$ times shorter than in the binary case. Consequently, the update strength of the probabilistic model has to be chosen $r$ times lower now. To investigate how desired model updates take place in this framework, we undertake a mathematical runtime analysis on the $r$-valued LeadingOnes problem. We prove that with the right parameters, the multi-valued UMDA solves this problem efficiently in $O(r\log(r)^2 n^2 \log(n))$ function evaluations. Overall, our work shows that EDAs can be adjusted to multi-valued problems, and it gives advice on how to set the main parameters.
翻译:显然,关于分配估算算法(EDAs)的所有研究都集中在假Boolean优化和变异问题上,因此我们采取初步步骤,在决定变量可能涉及两个以上但并非变异问题的问题上,使用EDAs。为此,我们提出一种自然的方法,将已知的单向分配算法(EDAs)扩大到这些变量。从天真到二进制情况,它避免了额外的限制。由于了解基因漂移对于最佳参数选择至关重要,我们将已知的遗传漂移定量分析扩大到多值变量的EDAs。粗略地说,当变量使用美元的不同值时,基因漂移变得显著的时间是二进制的两倍。因此,现在必须选择一种自然的方法,将已知的单向单向的 EADAs 的更新强度降低一倍。为了调查在这个框架内如何进行预期的模型更新,我们对估价值为$-EGEOs的问题进行了数学运行时间分析。我们证明,用正确的参数,多值的多值数据数据分析可以使我们的GADA2 能够有效地对整体价值进行评估。</s>