Andrew M. Sutton
ABSTRACT
The (1+1) EA is a simple evolutionary algorithm that is known to be efficient on linear functions and on some combinatorial optimization problems. In this paper, we rigorously study its behavior on two easy combinatorial problems: finding the 2-coloring of a class of bipartite graphs, and constructing satisfying assignments for a class of satisfiable 2-CNF Boolean formulas. We prove that it is inefficient on both problems in the sense that the number of iterations the algorithm needs to minimize the cost functions is superpolynomial with high probability. Our motivation is to better understand the influence of problem instance structure on the runtime character of a simple evolutionary algorithm. We are interested in what kind of structural features give rise to so-called metastable states at which, with probability 1 - o(1), the (1+1) EA becomes trapped and subsequently has difficulty leaving. Finally, we show how to modify the (1+1) EA slightly in order to obtain a polynomial-time performance guarantee on both problems.
- Bengt Apsvall, Michael F. Plass, and Robert Endre Tarjan. A linear time algorithm for testing the truth of certain quantified Boolean formulas. Information Processing Letters, 8(3):121--123, 1979.Google Scholar
Cross Ref
- Benjamin Doerr, Carola Doerr, and Franziska Ebel. Lessons from the black-box: Fast crossover-based genetic algorithms. In Proceedings of the Genetic and Evolutionary Computation Conference (GECCO), pages 781--788. ACM, 2013. Google ScholarDigital Library
- Benjamin Doerr, Daniel Johannsen, and Carola Winzen. Multiplicative drift analysis. Algorithmica, 64(4):673--697, 2012.Google ScholarDigital Library
- William Feller. Generalization of a probability limit theorem of Cramér. Transactions of the American Mathematical Society, 54(3):361--372, 1943.Google Scholar
- Oliver Giel and Ingo Wegener. Evolutionary algorithms and the maximum matching problem. In Helmut Alt and Michel Habib, editors, Proceedings of the Symposium on Theoretical Aspects of Computer Science (STACS), volume 2607 of Lecture Notes in Computer Science, pages 415--426. Springer, 2003. Google ScholarDigital Library
- Jiří Matoušek and Jan Vondrák. The probabilistic method (lecture notes), March 2008. http://kam.mff.cuni.cz/~matousek/prob-ln.ps.gz.Google Scholar
- Colin McDiarmid. A random recolouring method for graphs and hypergraphs. Combinatorics, Probability & Computing, 2:363--365, 1993.Google ScholarCross Ref
- Frank Neumann, Joachim Reichel, and Martin Skutella. Computing minimum cuts by randomized search heuristics. Algorithmica, 59(3):323--342, 2011. Google ScholarDigital Library
- Pietro S. Oliveto and Carsten Witt. Erratum: Simplified drift analysis for proving lower bounds in evolutionary computation. arXiv:1211.7184 {cs.NE}.Google Scholar
- Pietro S. Oliveto and Carsten Witt. Simplified drift analysis for proving lower bounds in evolutionary computation. Algorithmica, 59(3):369--386, 2011. Google ScholarDigital Library
- Christos H. Papadimitriou, Alejandro A. Schaffer, and Mihalis Yannakakis. On the complexity of local search. In Proceedings of the Annual ACM Symposium on Theory of Computing (STOC), pages 438--445. ACM, 1990. Google ScholarDigital Library
- Dirk Sudholt. Crossover is provably essential for the Ising model on trees. In Proceedings of the Genetic and Evolutionary Computation Conference (GECCO), pages 1161--1167. ACM, 2005. Google ScholarDigital Library
- David Williams. Probability with Martingales. Cambridge Mathematical Textbooks. Cambridge University Press, 1991.Google ScholarCross Ref
- Carsten Witt. Tight bounds on the optimization time of a randomized search heuristic on linear functions. Combinatorics, Probability and Computing, 22(2):294--318, 2013.Google ScholarCross Ref
Index Terms
- Superpolynomial lower bounds for the (1+1) EA on some easy combinatorial problems
Recommendations
Superpolynomial Lower Bounds for the $$(1+1)$$(1+1) EA on Some Easy Combinatorial Problems
The ($$1+1$$1+1) EA is a simple evolutionary algorithm that is known to be efficient on linear functions and on some combinatorial optimization problems. In this paper, we rigorously study its behavior on three easy combinatorial problems: finding the 2-...
Kernelization Lower Bounds Through Colors and IDs
In parameterized complexity, each problem instance comes with a parameter k, and a parameterized problem is said to admit a polynomial kernel if there are polynomial time preprocessing rules that reduce the input instance to an instance with size ...
Lower bounds on kernelization
Preprocessing (data reduction or kernelization) to reduce instance size is one of the most commonly deployed heuristics in the implementation practice to tackle computationally hard problems. However, a systematic theoretical study of them has remained ...
Comments