|
|
|
|
|
|
|
In the particular case of the interaction of crossover and inversion with reproduction a net of associations is induced (p. 108). Coadapted attributes (attributes defining schemata of above-average performance) become tightly linked and increase their proportion in the population (p. 127). In fact (p. 137) the expected rate of increase dPx/dt of the proportion of any given schema x is closely approximated by |
|
|
|
|
|
|
|
|
 |
|
|
|
|
|
|
|
|
where ax is the average excess of the random variable x in the population  . This formula is analogous to Fisher's (1930) classical result for single alleles and reduces to it when x is restricted to a single defining position. The resulting intrinsic parallelism greatly ameliorates the conflict between search and exploitation (obstacle (5)). By building up representations and models in terms of a language like the broadcast language (p. 152) the overall advantages of the schema approach can also be brought to bear on the problem of non-payoff information (obstacle (6)). The schemata provide for apportionment of credit to various aspects of the model on the basis of their relevance to realized predictions. |
|
|
|
|
|
|
|
|
At the time of this writing several computer studies of genetic plans have been completed (and more are underway). Four studies closely related to the theoretical framework will be outlined here: R. S. Rosenberg's Simulation of Genetic Populations with Biochemical Properties (1967), D. J. Cavicchio's Adaptive Search Using Simulated Evolution (1970), R. B. Hollstien's Artificial Genetic Adaptation in Computer Control Systems (1971), and D. R. Frantz's Non-linearities in Genetic Adaptive Search (1972). |
|
|
|
|
|
|
|
|
Richard Rosenberg completed his computer study of closed, small populations while formulation of the theoretical framework was still in its early stages. He concentrated on the complicated relationship between genotype and phenotype under dynamic interaction between the population and its environment. The model's central feature is the definition of phenotype by chemical concentrations. These concentrations are controlled by enzymes under genetic control. Epistasis has a critical role because several enzymes (and hence the corresponding genes) can affect any given phenotypic characteristic (chemical concentration). Though the variety of molecules, enzyme-controlled reactions, and genes is kept small to make the study feasible, it still presents a detailed picture of the propagation of |
|
|
|
|
|