On the analysis of the (1+ 1) evolutionary algorithm
Theoretical Computer Science
On the Choice of the Mutation Probability for the (1+1) EA
PPSN VI Proceedings of the 6th International Conference on Parallel Problem Solving from Nature
Probability and Computing: Randomized Algorithms and Probabilistic Analysis
Probability and Computing: Randomized Algorithms and Probabilistic Analysis
Upper and Lower Bounds for Randomized Search Heuristics in Black-Box Optimization
Theory of Computing Systems
Automated Unique Input Output Sequence Generation for Conformance Testing of FSMs
The Computer Journal
On the impact of the mutation-selection balance on the runtime of evolutionary algorithms
Proceedings of the tenth ACM SIGEVO workshop on Foundations of genetic algorithms
PPSN'10 Proceedings of the 11th international conference on Parallel problem solving from nature: Part I
Optimizing monotone functions can be difficult
PPSN'10 Proceedings of the 11th international conference on Parallel problem solving from nature: Part I
PPSN'10 Proceedings of the 11th international conference on Parallel problem solving from nature: Part I
ICONIP'11 Proceedings of the 18th international conference on Neural Information Processing - Volume Part II
The max problem revisited: the importance of mutation in genetic programming
Proceedings of the 14th annual conference on Genetic and evolutionary computation
Parsimony pressure versus multi-objective optimization for variable length representations
PPSN'12 Proceedings of the 12th international conference on Parallel Problem Solving from Nature - Volume Part I
Hi-index | 0.00 |
Many practical optimisation problems allow candidate solutions of varying lengths, and where the length of the optimal solution is thereby a priori unknown. We suggest that non-uniform mutation rates can be beneficial when solving such problems. In particular, we consider a mutation operator that flips each bit with a probability that is inversely proportional to the bit position, rather than the bitstring length. The runtime of the (1+1) EA using this mutation operator is analysed rigorously on standard example functions. Furthermore, the behaviour of the new mutation operator is investigated empirically on a real world software engineering problem that has variable, and unknown solution lengths. The results show how the speedup that can be achieved with the new operator depends on the distribution of the solution lengths in the solution space. We consider a truncated geometric distribution, and show that the new operator can yield exponentially faster runtimes for some parameters of this distribution. The experimental results show that the new mutation operator leads to dramatically shorter runtimes on a class of instances of the software engineering problem that is conjectured to have short solutions on average.