Generating hard satisfiability problems
Artificial Intelligence - Special volume on frontiers in problem solving: phase transitions and complexity
Constructing an asymptotic phase transition in random binary constraint satisfaction problems
Theoretical Computer Science - Phase transitions in combinatorial problems
Generating Satisfiable Problem Instances
Proceedings of the Seventeenth National Conference on Artificial Intelligence and Twelfth Conference on Innovative Applications of Artificial Intelligence
Random Constraint Satisfaction: Theory Meets Practice
CP '98 Proceedings of the 4th International Conference on Principles and Practice of Constraint Programming
Many hard examples in exact phase transitions
Theoretical Computer Science
Evolving combinatorial problem instances that are difficult to solve
Evolutionary Computation
Random constraint satisfaction: Easy generation of hard (satisfiable) instances
Artificial Intelligence
Hiding satisfying assignments: two are better than one
AAAI'04 Proceedings of the 19th national conference on Artifical intelligence
Exact phase transitions in random constraint satisfaction problems
Journal of Artificial Intelligence Research
Where the really hard problems are
IJCAI'91 Proceedings of the 12th international joint conference on Artificial intelligence - Volume 1
AAAI'96 Proceedings of the thirteenth national conference on Artificial intelligence - Volume 1
On the phase transitions of random k-constraint satisfaction problems
Artificial Intelligence
Heuristic techniques for variable and value ordering in CSPs
Proceedings of the 13th annual conference on Genetic and evolutionary computation
Hi-index | 0.00 |
In computer science it is a common practice to evaluate the performance of algorithms using a set of benchmark or randomly generated instances. However, following that approach, the weaknesses of the algorithms may not be exposed. This work is the first phase of research project on coevolution of solutions methods versus problem instances. The goal of study is to generate a method to find difficult to solve problem instances capable of challenging the solution methods or algorithms under analysis, helping to discover opportunities for improvement. An evolutionary model is proposed to find hard binary constraint satisfaction problem instances for different variable ordering heuristics. We characterize the search space by generating random instances with different values for the constraint density and tightness. For all the heuristics, the most difficult problems are located in the same region of the space near to the phase transition. However, there are certain regions of the search space where a heuristic dominates the others, especially where the problems are solvable. Finally, we compare the hardest instances found during the search space exploration with the outcome instances of the evolutionary model. The results show that evolved instances are harder to solve than the ones randomly generated.