<p>The <InlineEquation ID="IEq5"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="453_2025_1323_Article_IEq5.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="52" /> </InlineMediaObject> <EquationSource Format="TEX">\(\textsc {Jump} _k\)</EquationSource> </InlineEquation> benchmark was the first problem for which crossover was proven to give a speed-up over mutation-only evolutionary algorithms. Jansen and Wegener (Algorithmica 2002) proved an upper bound of <InlineEquation ID="IEq6"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="453_2025_1323_Article_IEq6.gif" Format="GIF" Height="21" Rendition="HTML" Resolution="72" Type="Linedraw" Width="137" /> </InlineMediaObject> <EquationSource Format="TEX">\(O(\textrm{poly}(n) + 4^k/p_c)\)</EquationSource> </InlineEquation> for the (<InlineEquation ID="IEq7"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="453_2025_1323_Article_IEq7.gif" Format="GIF" Height="12" Rendition="HTML" Resolution="72" Type="Linedraw" Width="15" /> </InlineMediaObject> <EquationSource Format="TEX">\(\mu \)</EquationSource> </InlineEquation>+1)&#xa0;Genetic Algorithm ((<InlineEquation ID="IEq8"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="453_2025_1323_Article_IEq7.gif" Format="GIF" Height="12" Rendition="HTML" Resolution="72" Type="Linedraw" Width="15" /> </InlineMediaObject> <EquationSource Format="TEX">\(\mu \)</EquationSource> </InlineEquation>+1)&#xa0;GA), but only for unrealistically small crossover probabilities&#xa0;<InlineEquation ID="IEq9"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="453_2025_1323_Article_IEq9.gif" Format="GIF" Height="12" Rendition="HTML" Resolution="72" Type="Linedraw" Width="18" /> </InlineMediaObject> <EquationSource Format="TEX">\(p_c\)</EquationSource> </InlineEquation>. To this date, it remains an open problem to prove similar upper bounds for realistic&#xa0;<InlineEquation ID="IEq10"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="453_2025_1323_Article_IEq9.gif" Format="GIF" Height="12" Rendition="HTML" Resolution="72" Type="Linedraw" Width="18" /> </InlineMediaObject> <EquationSource Format="TEX">\(p_c\)</EquationSource> </InlineEquation>; the best known runtime bound, in terms of function evaluations, for <InlineEquation ID="IEq11"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="453_2025_1323_Article_IEq11.gif" Format="GIF" Height="19" Rendition="HTML" Resolution="72" Type="Linedraw" Width="70" /> </InlineMediaObject> <EquationSource Format="TEX">\(p_c = \Omega (1)\)</EquationSource> </InlineEquation> is <InlineEquation ID="IEq12"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="453_2025_1323_Article_IEq12.gif" Format="GIF" Height="21" Rendition="HTML" Resolution="72" Type="Linedraw" Width="91" /> </InlineMediaObject> <EquationSource Format="TEX">\(O((n/\chi )^{k-1})\)</EquationSource> </InlineEquation>, <InlineEquation ID="IEq13"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="453_2025_1323_Article_IEq13.gif" Format="GIF" Height="12" Rendition="HTML" Resolution="72" Type="Linedraw" Width="15" /> </InlineMediaObject> <EquationSource Format="TEX">\(\chi \)</EquationSource> </InlineEquation> a positive constant. We provide a novel approach and analyse the evolution of the population diversity, measured as sum of pairwise Hamming distances, for a variant of the (<InlineEquation ID="IEq14"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="453_2025_1323_Article_IEq7.gif" Format="GIF" Height="12" Rendition="HTML" Resolution="72" Type="Linedraw" Width="15" /> </InlineMediaObject> <EquationSource Format="TEX">\(\mu \)</EquationSource> </InlineEquation>+1)&#xa0;GA on <InlineEquation ID="IEq15"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="453_2025_1323_Article_IEq5.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="52" /> </InlineMediaObject> <EquationSource Format="TEX">\(\textsc {Jump} _k\)</EquationSource> </InlineEquation>. The (<InlineEquation ID="IEq16"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="453_2025_1323_Article_IEq7.gif" Format="GIF" Height="12" Rendition="HTML" Resolution="72" Type="Linedraw" Width="15" /> </InlineMediaObject> <EquationSource Format="TEX">\(\mu \)</EquationSource> </InlineEquation>+1)-<InlineEquation ID="IEq17"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="453_2025_1323_Article_IEq17.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="18" /> </InlineMediaObject> <EquationSource Format="TEX">\({\lambda _c}\)</EquationSource> </InlineEquation>-GA creates one offspring in each generation either by applying mutation to one parent or by applying crossover <InlineEquation ID="IEq18"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="453_2025_1323_Article_IEq17.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="18" /> </InlineMediaObject> <EquationSource Format="TEX">\({\lambda _c}\)</EquationSource> </InlineEquation> times to the same two parents (followed by mutation), to amplify the probability of creating an accepted offspring in generations with crossover. We show that population diversity in the (<InlineEquation ID="IEq19"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="453_2025_1323_Article_IEq7.gif" Format="GIF" Height="12" Rendition="HTML" Resolution="72" Type="Linedraw" Width="15" /> </InlineMediaObject> <EquationSource Format="TEX">\(\mu \)</EquationSource> </InlineEquation>+1)-<InlineEquation ID="IEq20"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="453_2025_1323_Article_IEq17.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="18" /> </InlineMediaObject> <EquationSource Format="TEX">\({\lambda _c}\)</EquationSource> </InlineEquation>-GA converges to an equilibrium of near-perfect diversity. This yields an improved time bound of <InlineEquation ID="IEq21"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="453_2025_1323_Article_IEq21.gif" Format="GIF" Height="21" Rendition="HTML" Resolution="72" Type="Linedraw" Width="128" /> </InlineMediaObject> <EquationSource Format="TEX">\(O(\mu n \log (\mu ) + 4^k)\)</EquationSource> </InlineEquation> function evaluations for a range of&#xa0;<i>k</i> under the mild assumptions <InlineEquation ID="IEq22"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="453_2025_1323_Article_IEq22.gif" Format="GIF" Height="19" Rendition="HTML" Resolution="72" Type="Linedraw" Width="90" /> </InlineMediaObject> <EquationSource Format="TEX">\(p_c = O(1/k)\)</EquationSource> </InlineEquation> and <InlineEquation ID="IEq23"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="453_2025_1323_Article_IEq23.gif" Format="GIF" Height="19" Rendition="HTML" Resolution="72" Type="Linedraw" Width="77" /> </InlineMediaObject> <EquationSource Format="TEX">\(\mu \in \Omega (kn)\)</EquationSource> </InlineEquation>. For all constant&#xa0;<i>k</i>, the restriction is satisfied for some <InlineEquation ID="IEq24"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="453_2025_1323_Article_IEq11.gif" Format="GIF" Height="19" Rendition="HTML" Resolution="72" Type="Linedraw" Width="70" /> </InlineMediaObject> <EquationSource Format="TEX">\(p_c = \Omega (1)\)</EquationSource> </InlineEquation> and it implies that the expected runtime for all constant&#xa0;<i>k</i> and an appropriate <InlineEquation ID="IEq25"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="453_2025_1323_Article_IEq25.gif" Format="GIF" Height="19" Rendition="HTML" Resolution="72" Type="Linedraw" Width="79" /> </InlineMediaObject> <EquationSource Format="TEX">\(\mu = \Theta (kn)\)</EquationSource> </InlineEquation> is bounded by <InlineEquation ID="IEq26"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="453_2025_1323_Article_IEq26.gif" Format="GIF" Height="20" Rendition="HTML" Resolution="72" Type="Linedraw" Width="80" /> </InlineMediaObject> <EquationSource Format="TEX">\(O(n^2 \log n)\)</EquationSource> </InlineEquation>, irrespective of&#xa0;<i>k</i>. For larger&#xa0;<i>k</i>, the expected time of the (<InlineEquation ID="IEq27"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="453_2025_1323_Article_IEq7.gif" Format="GIF" Height="12" Rendition="HTML" Resolution="72" Type="Linedraw" Width="15" /> </InlineMediaObject> <EquationSource Format="TEX">\(\mu \)</EquationSource> </InlineEquation>+1)-<InlineEquation ID="IEq28"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="453_2025_1323_Article_IEq17.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="18" /> </InlineMediaObject> <EquationSource Format="TEX">\({\lambda _c}\)</EquationSource> </InlineEquation>-GA is <InlineEquation ID="IEq29"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="453_2025_1323_Article_IEq29.gif" Format="GIF" Height="21" Rendition="HTML" Resolution="72" Type="Linedraw" Width="41" /> </InlineMediaObject> <EquationSource Format="TEX">\(\Theta (4^k)\)</EquationSource> </InlineEquation>, which is tight for a large class of unbiased black-box algorithms and faster than the original (<InlineEquation ID="IEq30"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="453_2025_1323_Article_IEq7.gif" Format="GIF" Height="12" Rendition="HTML" Resolution="72" Type="Linedraw" Width="15" /> </InlineMediaObject> <EquationSource Format="TEX">\(\mu \)</EquationSource> </InlineEquation>+1)&#xa0;GA by a factor of <InlineEquation ID="IEq31"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="453_2025_1323_Article_IEq31.gif" Format="GIF" Height="19" Rendition="HTML" Resolution="72" Type="Linedraw" Width="56" /> </InlineMediaObject> <EquationSource Format="TEX">\(\Omega (1/p_c)\)</EquationSource> </InlineEquation>. We also show that our analysis can be extended to other unitation functions such as <InlineEquation ID="IEq32"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="453_2025_1323_Article_IEq32.gif" Format="GIF" Height="19" Rendition="HTML" Resolution="72" Type="Linedraw" Width="61" /> </InlineMediaObject> <EquationSource Format="TEX">\(\textsc {Jump} _{k, \delta }\)</EquationSource> </InlineEquation> and H<span>urdle</span>.</p>

错误:搜索内容不能为空,请输入英文关键词
错误:关键词超出字数限制,请精简
高级检索

Achieving Tight \(O(4^k)\) Runtime Bounds on Jumpk by Proving that Genetic Algorithms Evolve Near-Maximal Population Diversity

  • Andre Opris,
  • Johannes Lengler,
  • Dirk Sudholt

摘要

The \(\textsc {Jump} _k\) benchmark was the first problem for which crossover was proven to give a speed-up over mutation-only evolutionary algorithms. Jansen and Wegener (Algorithmica 2002) proved an upper bound of \(O(\textrm{poly}(n) + 4^k/p_c)\) for the ( \(\mu \) +1) Genetic Algorithm (( \(\mu \) +1) GA), but only for unrealistically small crossover probabilities  \(p_c\) . To this date, it remains an open problem to prove similar upper bounds for realistic  \(p_c\) ; the best known runtime bound, in terms of function evaluations, for \(p_c = \Omega (1)\) is \(O((n/\chi )^{k-1})\) , \(\chi \) a positive constant. We provide a novel approach and analyse the evolution of the population diversity, measured as sum of pairwise Hamming distances, for a variant of the ( \(\mu \) +1) GA on \(\textsc {Jump} _k\) . The ( \(\mu \) +1)- \({\lambda _c}\) -GA creates one offspring in each generation either by applying mutation to one parent or by applying crossover \({\lambda _c}\) times to the same two parents (followed by mutation), to amplify the probability of creating an accepted offspring in generations with crossover. We show that population diversity in the ( \(\mu \) +1)- \({\lambda _c}\) -GA converges to an equilibrium of near-perfect diversity. This yields an improved time bound of \(O(\mu n \log (\mu ) + 4^k)\) function evaluations for a range of k under the mild assumptions \(p_c = O(1/k)\) and \(\mu \in \Omega (kn)\) . For all constant k, the restriction is satisfied for some \(p_c = \Omega (1)\) and it implies that the expected runtime for all constant k and an appropriate \(\mu = \Theta (kn)\) is bounded by \(O(n^2 \log n)\) , irrespective of k. For larger k, the expected time of the ( \(\mu \) +1)- \({\lambda _c}\) -GA is \(\Theta (4^k)\) , which is tight for a large class of unbiased black-box algorithms and faster than the original ( \(\mu \) +1) GA by a factor of \(\Omega (1/p_c)\) . We also show that our analysis can be extended to other unitation functions such as \(\textsc {Jump} _{k, \delta }\) and Hurdle.