Biologically motivated schemes [for combinatorial optimization]

The process of biological evolution by natural selection can be thought of as an iterative procedure for optimization. Usually, however, what is being optimized is some aspect of the form or behavior of an organism, which represents a very complicated constraint on the underlying genetic material. (It is as if one is defining constraints on the initial conditions for a cellular automaton by looking at the pattern generated by the cellular automaton after a long time.) But the strategies of biological evolution can also be used in trying to satisfy simpler constraints. Two of the most important strategies are maintaining a whole population of individuals, not just the single best result so far, and using sex to produce large-scale mixing. But once again, while these strategies may in some cases lead to greater efficiency, they do not usually lead to qualitative differences. (See also page 1105.)