|
|
|
|
|
|
|
to each individual, t1 acts as follows: At the beginning of each time period t, the plan's accumulated information about the environment resides in a finite population  selected from  . The most important part of this information is given by the discrete distributions which give the proportions of different sets of alleles in the population . serves not only as the plan's repository of accumulated information, but also as the source of new variants which will give rise to  . As indicated earlier, the formation of proceeds in two phases. During the first phase, is modified to form  by copying each individual in a number of times dependent upon the individual's observed performance. The number of copies made will be determined stochastically so that the expected number of copies increases in proportion to observed performance. During the second phase, the operators are applied to the population , interchanging and modifying the sets of alleles, to produce the new generation |
|
|
|
|
|
|
|
|
One key to understanding t1's resolution of the dilemma lies in observing what happens to small sets of adjacent alleles under its action. In particular, what happens if an adjacent set of alleles appears in several different chromosomes of above-average fitness and not elsewhere? Because each of the chromosomes will be duplicated an above-average number of times, the given alleles will occupy an increased proportion of the population after the duplication phase. This increased proportion will of course result whether or not the alleles had anything to do with the above-average fitness. The appearance of the alleles in the extra-fit chromosomes might be happenstance, but it is equally true that any correlation between the given selection of alleles and above-average fitness will be exploited by this action. Moreover, the more varied the chromosomes containing the alleles, the less likely it is that the alleles and above-average fitness are uncorrelated. |
|
|
|
|
|
|
|
|
What happens now when the genetic operators W are applied to form the next generation? As indicated earlier, the closer alleles are to one another in the chromosome the less likely they are to be separated during the operator phase. Thus the operator phase usually transfers adjacent sets of genes as unit, placing them in new chromosomal contexts without disturbing them otherwise. These new contexts further test the sets of alleles for correlation with above-average fitness. If the selected set of alleles does indeed augment fitness, the chromosomes containing the set will again (on the average) be extra fit. On the other hand, if the prior associations were simply happenstance, sustained association with extra-fit chromosomes becomes increasingly less likely as the number of trials (new contexts) increases. The net effect of the genetic plan over several generations will be an increasing predominance of alleles and sets of alleles augmenting fitness in the given environment. |
|
|
|
|
|