Lamarckian Evolution , The Baldwin E ectand Function

We compare two forms of hybrid genetic search. The rst uses Lamarckian evolution, while the second uses a related method where local search is employed to change the tness of strings, but the acquired improvements do not change the genetic encoding of the individual. The latter search method exploits the Baldwin eeect. By modeling a simple genetic algorithm we show that functions exist where simple genetic algorithms without learning as well as Lamar-ckian evolution converge to the same local optimum, while genetic search utilizing the Baldwin eeect converges to the global optimum. We also show that a simple genetic algorithm exploiting the Baldwin eeect can sometimes outperform forms of Lamarckian evolution that employ the same local search strategy.