|
|
|
|
|
|
|
rate. In genetics the decreases are provided by deletion. The easiest generalization of these operators is an operator which doubles (or halves) the number of copies at a randomly chosen (set of adjacent) location(s). |
|
|
|
|
|
|
|
|
Though the operators just described are useful, they are not necessary. Moreover they do not compensate the major shortcoming of genetic plans which use just the first three operators described. That shortcoming is the complete dependence of such plans upon the detectors determining the representation. If the set of detectors {di}is inadequate, in any way, the plan must operate within that constraint. However, if the plan could add or modify detectors at need, it could circumvent the difficulty. This implies making the detectors themselves subject to adaptation. When we note that each detector can be specified by an appropriate subroutine (string of instructions) for a general purpose computer, a way of making this extension suggests itself. By keeping the number of basic instructions from which the subroutines are constructed small, we can treat them as alleles.  can then be extended to include all strings of basic instructions. In this way contains a representation of any possible detector, set of detectors or, in fact, any effectively describable way of processing information. Moreover, under this extension, favored schemata correspond to useful coordinated sets of instructions (such as detectors). Genetic plans applied to , so extended, can thus develop whatever functions or representations they need. This problem and the suggested approach are complex enough to merit a chapter, chapter 8. |
|
|
|
|
|
|
|
|
The Jacob-Monod (1961) "operon" model of the functioning of the chromosome has an interesting relation to the extension of just suggested. In the extension, we can think of each element of as a program processing inputs from the environment to produce outputs affecting that environment (cf. chapter 3.4 where transformations {hi} are the outputs). The performance of the element is thus directly determined by the relevance or fitness of the program. The "operon" model treats the chromosome as a similar information processing device. Each gene can either be active (cf. the execution of an instruction) or inactive. When active the gene is participating in the production of signals (enzymes) which modulate the ongoing activity of the cell. It thereby determines the cell's modes of action and critical aspects of its structure. The genes are collected in groupsoperonssuch that all genes in the group are either simultaneously active or inactive, as determined by one control gene in the group called an ''operator gene" (or more recently, a "receptor gene" in Britten and Davidson 1969; see Fig. 14). The remainder of the cell is treated as the chromosome's environment. The action of the "receptor gene" is conditional upon the presence of signals (proteins) from the cell (usually through the mediation of other genes"repressor" or "sensor" |
|
|
|
|
|