When Feasible Solutions Become a Trap: Initialization Strategies for EDAs in the Traveling Tournament Problem
tags: Traveling Tournament Problem, Estimation of Distribution Algorithms@GuilhermeNakahata
I’ll be presenting this paper at ECTA 2026 in Angers, France, on October 28–30. It looks at how different starting solutions affect Estimation of Distribution Algorithms (EDAs) for the Traveling Tournament Problem.
We compared feasible, partially feasible, and infeasible initial populations using a univariate EDA and a tree-based EDA. We also studied the effects of penalty strength and local search, our results showed that initialization had more influence on the search than penalty strength.
Figure generated by Daan van den Berg.
The full paper can be seen here
You can also read our earlier work on generating feasible initial schedules for the TTP here.