Generalized Cases: Representation and Steps Towards Efficient Similarity Assessment

For certain application areas of case-based reasoning, the traditional view of cases as points in the problem-solution space is not appropriate. Motivated by a concrete application in the area of electronic design reuse, we introduce the concept of a generalized case that represents experience that naturally covers a space rather than a point. Within a formal framework we introduce the semantics of generalized cases and derive a canonical similarity measure for them. Generalized cases can be represented in a very flexible way by using constraints. This representation asks for new means of similarity assessment. We argue that in principle fuzzy constraint satisfaction or non-linear programming can be applied for similarity computation. However, to avoid the computational complexity of these approaches, we propose an algorithm for an efficient estimation of similarity for generalized cases.