Yazar "Corus, Dogan" seçeneğine göre listele

Listeleniyor 1 - 3 / 3
  • Küçük Resim Yok
    Öğe
    Fast Contiguous Somatic Hypermutations for Single-Objective Optimisation and Multi-Objective Optimisation Via Decomposition
    (Assoc Advancement Artificial Intelligence, 2025) Corus, Dogan; Oliveto, Pietro S.; Yazdani, Donya
    Somatic Contiguous Hypermutations (CHM) are a popular variation operator used in artificial immune systems for optimisation tasks. Theoretical studies have shown that CHM operators can lead to considerable speed-ups in the expected optimisation time compared to the traditional standard bit mutation (SBM) operators used in evolutionary computation for both single-objective and multi-objective problems where it is advantageous to mutate large contiguous areas of the genotype representing the candidate solutions. These speed-ups can make the difference between polynomial and exponential runtimes, but come at the expense of the CHM operator being considerably slower than the SBM operator in easy hillclimbing phases of the optimisation process, when small areas of the genotype have to be mutated for progress to be made. In this paper we present a Fast CHM operator that is asymptotically just as fast as traditional SBM for hillclimbing yet maintains the efficacy of the standard CHM operator when large jumps in the search space are required to make progress efficiently. We demonstrate such efficacy on all applications where CHM has been previously studied in the literature.
  • Küçük Resim Yok
    Öğe
    Hybrid Selection Allows Steady-State Evolutionary Algorithms to Control the Selective Pressure in Multimodal Optimisation
    (Assoc Computing Machinery, 2025) Corus, Dogan; Oliveto, Pietro S.; Zheng, Feiyang
    Recent work has shown that Inverse Tournament Selection operators within steady-state evolutionary algorithms (EAs) allow to control the selective pressure much more accurately than in generational EAs. However, to achieve low selective pressures, large tournament sizes are required which come at the cost of prohibitive expected times for the population to escape from local optima. To this end, we propose a hybrid selection mechanism that leads to considerable speed-ups in the expected time to escape from local optima while permitting to keep the selective pressure arbitrarily low and the use of large population sizes. The mechanism simply switches between Inverse Elitist selection and Uniform selection when it detects that the population is stuck on local optima, and switches back when an improving solution is found. We prove its effectiveness for the TruncatedTwomax.. and RidgeWithBranches.. benchmarks from the literature by providing super-linear speed-ups over the (.. +1) EA with any fixed selective pressure.
  • Küçük Resim Yok
    Öğe
    On the Generalisation Performance of Geometric Semantic Genetic Programming for Boolean Functions: Learning Block Mutations
    (Assoc Computing Machinery, 2025) Corus, Dogan; Oliveto, Pietro S.
    In this paper we present the first rigorous theoretical analysis of the generalisation performance of a Geometric Semantic Genetic Programming (GSGP) system. More specifically, we consider a hill-climber using the GSGP Fixed Block Mutation (FBM) operator for the domain of Boolean functions. We prove that the algorithm cannot evolve Boolean conjunctions of arbitrary size that are correct on unseen inputs chosen uniformly at random from the complete truth table i.e., it generalises poorly. Two algorithms based on the Varying Block Mutation (VBM) operator are proposed and analysed to address the issue. We rigorously prove that under the uniform distribution the first one can efficiently evolve any Boolean function of constant size with respect to the number of available variables, while the second one can efficiently evolve general conjunctions or disjunctions of any size without requiring prior knowledge of the target function class. An experimental analysis confirms the theoretical insights for realistic problem sizes and indicates the superiority of the proposed operators also for small parity functions not explicitly covered by the theory.