next up previous

4.3.5 Convergence Criteria

Two standard convergence tests are used to terminate ES searches. One is that the absolute difference in the objective function values of the best and worst members of the post-selection population is less than a user-specified limit, i.e.

The other is that the relative difference in the objective function values of the best and worst members of the post-selection population is less than a user-specified limit, i.e.

Thus, absolutely or relatively, the objective function values of the parents must lie close together before the algorithm is deemed to have converged.