Enough of the theoretical framework has now been erected that we can begin to view it as a whole. To this end, the present chapter will discuss three general aspects of the theory. Section 1 will concentrate on those insights offered by the theory which are useful across the full spectrum of adaptive problems. Section 2 provides a synopsis of several computer studies to give the reader an idea of how the overall theory works in particular contexts. Section 3 will outline several difficult long-range problems which fall within the scope of the theory.
1. Insights
Within the theoretical framework problems of adaptation have been phrased in terms of generating structures of progressively higher performance. Because the framework itself places no constraints on what objects can be taken as structures, other than that it be possible to rank them according to some measure of performance, the resulting theory has considerable latitude. Once adaptation has been characterized along these lines, it is also relatively easy to describe several pervasive, interrelated obstacles to adaptationobstacles which occur in some combination in all but the most trivial problems:
1. High cardinality of . The set of potentially interesting structures is extensive, making searches long and storage of relevant data difficult.
2. Apportionment of credit. Knowledge of properties held in common by structures of above-average performance is incomplete, making it difficult to infer from past tests what untested structures are likely to yield above-average performance.
3. High dimensionality of µE. Performance is a function of large numbers of variables, making it difficult to use classical optimization methods employing gradients, etc.
. The set of potentially interesting structures is extensive, making searches long and storage of relevant data difficult.