[W]hen using mutation at a fairly high rate the lower salient genes do not fully converge since the selective pressure acting on these genes is counterbalanced by the disruptive effect of the mutation operator. – Thierens (1998)1
“Drift Stall” or “Convergence Stall” can result from overly high rates of mutation. Thierens (1998)1 credited Rudnick (1992)2 for the observation that where the mutation rate is very high, lower salient genes may be prevented from converging at all, due to the level of noise created by the mutation operator.
The key factor is to ensure that the diversity of the less salient building blocks is preserved long enough in order to buy time for the selection-recombination operators to reach this level of fitness scale. – Thierens (1998)1
- Domino Convergence, Drift, and the Temporal-Salience Structure of Problems – Dirk Thierens, David E. Goldberg, Angela Guimaraes Pereira, 1998
- Rudnick M. (1992). Genetic Algorithms and Fitness Variance with an Application to the Automated Design of Artificial Neural Networks. PhD thesis, Oregon Graduate Institute of Science and Technology.