Natural processes which accelerate the evolutionary search
Wirt Atmar · 2005
Natural evolution is an iterative optimization process such that a population of trials is subject to repetitive selection, a process which operates to minimize functional (behavioral) error. The physics employed in common numeric search techniques is structurally different than that characteris- tic of natura1 evolutionary processes. Sufficient experience has now been garnered to suggest that all strongly-determined numerical optimization methods tend to work acceptably well on some adaptive surfaces, but exhibit a pronounced tendency to stall in indefinite oscillation, fail to converge, or become entrapped in local optima on others. These stalls are caused by the highly-determined correlation between parent and child trial vectors. But this is not the manner in which evolutionary optimization proceeds. The mutagenesis guaranteed by repli- cative error generates a continuum of fine- and large-grained mutations. The correlation between parent and child trial vec- tors is often strong, but it is never absolute. Under random mu- tation, no combination is impossible, thus global solutions in a finite state space are guaranteed in infinite time. Natural- ly-occurring processes which accelerate this search are dis- cussed using two simple models, one combinatorial, the orher time-sequential.