Effective Adaptive Plans A Hypothetical Search Process
Kees P. Pieters · 2006
Many iterative search processes, or adaptive plans, that aim to find an optimal solution in a given problem domain, suggest that an optimal search process has an exponential character. Plans that consist of multiple strategies running in parallel, such as bandit searches, aim to demonstrate this pattern in probabilistic distributions of finding a best observed strategy amongst a number of alternatives. This paper introduces a hypothetical adaptive plan that consists of three strategies. One strategy guarantees a better result with each iteration, one has comparable results, and one guarantees worse results. The idea behind this approach is the suspicion that every adaptive plan can basically be mapped to these three base strategies, and that the exponential character of an optimal plan is a trait of its recurrent character.