|
|
|
|
|
|
|
competition in the absence of environmental payoff, the probability of such an event increasing as a function of elapsed time since last payoff. |
|
|
|
|
|
|
|
|
Much more detail would have to be supplied for this implementation of predictive modeling to reach the level of precision earlier given to the description of genetic plans. However the objective here is only to indicate the potential of the broadcast language for predictive modeling with changing representations. |
|
|
|
|
|
|
|
|
As a final example, note that the world of radiative signals (sound, light, etc.) is susceptible to modeling as a complex broadcast system. In fact one physical realization of devices specified in the broadcast language would assign a unique frequency to each signal and realize broadcast units as a variety of frequency modulation devices. |
|
|
|
|
|
|
|
|
Even where the broadcast language does not so directly represent extant models, it still supplies a rigorous framework for the description and modification of representations. In particular it makes possible the application of genetic plans to the problem of discovering suitable representations. Because devices are represented by strings and because the functional elements (the broadcast units) are self-defined, the generalized genetic operators of sections 6.2 and 6.3 can be used to modify the devices. Moreover, as indicated in section 3, these operators can themselves be defined within the broadcast language. This makes possible a hierarchy of operators defined with respect to various kinds of punctuation. Thus, one crossover operator could be defined to produce crossing-over anywhere along the string, another could be defined to produce crossing-over only at the symbol: (thereby providing for exchange of arguments between broadcast units), still another only at * (thereby exchanging broadcast units between devices), and so on. The operators so defined introduce a hierarchy of schemata ranging from schemata concerned primarily with varieties of arguments, through schemata concerned with combinations of broadcast units, and on to higher levels of organization (e.g., behavioral atoms, behavioral units, etc.). |
|
|
|
|
|
|
|
|
Note that for the broadcast language schemata are generally defined with respect to sets of arbitrarily long strings. That is, the set of all devices specifiable in the broadcast language would be the set of all strings which can be formed from the ten basic symbols; since a schema designates the set of all strings which match it on its defining positions, each schema designates a countable subset of devices. Using this extension of the notion of a schema, we see that the results of chapters 6 and 7, particularly those pertaining to intrinsic parallelism, extend directly to the adaptation of codings and representations. Since the operators themselves can also be specified within the broadcast language, they can also be made subject to the same adaptive processes. |
|
|
|
|
|