I read a few papers on Memetic Algorithms recently. The idea of meme as a unit of information that reproduces itself as people exchange idea is very interesting to me. When I review the brief history of my though during these recent years, I found out that I had a similar idea about 4 years ago: an ALife setting in which individual learns and then transfer their knowledge to each other when they become close (I always like spatially-extended ALife creatures that can move in the environment). However in my opinion, the idea of meme is somehow misused or biased in the evolutionary computation community. Most of them interpret a meme as a hybridization of local search and a genetic global search. In other words, most researchers use a kind of local search after (or before) each genetic operator they made. Of course, there are many variations in the exact implementation â but most of them are actually doing so. What is the root of this trend? I am not sure, may be it is due to Moscatoâs paper that shaped a new paradigm and getting out of it is difficult. But, I think it comes from this interpretation of meme: when people get idea, they change it during their lifetime, and then transfer it to others. Traditional genetic search does this transfer, but does not include a piece of âlifetime adaptationâ. They interpreted this lifetime adaptation as a kind of local search and most of them followed a Lamarckian approach as âmemes must be changed during its life in the individualâ.
Well â¦ I will write more about memetic algorithms.
(*): This quotation is actually a meme. Many papers refer to it when they want to define a meme. I guess only a few people actually read the Dawkins’ book. One of them is me!