next up previous
Next: 4.2 Methods for the Up: 4 Configuration Task Previous: 4 Configuration Task

4.1 Automated Configuration of Problem Solvers as Parametric Design

In this section we map the automated configuration of PSMs on the configuration task. In order to make this mapping, we consider the general characteristics of the configuration task given above in the context of the construction of problem solvers and we consider the configuration of PSMs as a search problem.

We first consider the input types of the configuration task in the context of configuring PSMs. The inputs are:

The output of the configuration of methods consists of the six components of particular types which are structured in such a way that together they represent a diagnostic method.

The three types of configurations (possible, valid, and suitable, Figure 2) can be given a meaning in configuring methods. A possible configuration is a method that contains a definition for each component of the general method schema. A valid configuration is a method that expresses a diagnostic method and has no conflicts with the assumptions under which the method must operate. A suitable configuration is a method that satisfies the desired goals.

The mapping from the elements of a general construction problem onto our problem of method construction shows that we can indeed interpret automated configuration of diagnostic problem solvers as a configuration problem. In fact, it can even be interpreted as parametric design, because we use a fixed structure and the possible definitions of each component can be considered as the range of the parameters in formula (1). However, in our view of configuring PSMs we do not only modify the method, but possibly also the assumptions, goals, and the input problem, as already stated in Section 2.



next up previous
Next: 4.2 Methods for the Up: 4 Configuration Task Previous: 4 Configuration Task



Frank van Harmelen
Fri Oct 4 13:40:35 MET DST 1996