Lattices of Knowledge in Intelligent Systems Validation
Klaus P. Jantke, Jörg Herrmann · elib (German Aerospace Center) · 1999
Validation is art of investigating whether or not a given system is the right according to users’ expectations and to necessities of application domain. Researchers all over world are currently striving hard to transform intelligent systems validation into a science. The focus of present paper is rather narrow. It alms at a very little contribution to advance formal background of systems validation. The authors hope that, even in case they do not complete succeed in their efforts towards a science of intelligent systems validation, their contribution might advance art a little. The paper deals with problems concerning knowledge sources which have to be utilized in systems validation. There is a particular focus on knowledge about system under inspection. The key technical term of this publication is lest case and key computational process is reduction of sets of test cases to subsets of data which axe still sufficiently expressive, but feasible. The crucial theoretical concept underlying this approach is inheritance of validity and key perspective which is underlying knowledge processing approach is circumscribed by concept of a lattice o] knowledge. As one cam usually build Boolean combinations of knowledge units, i.e. one may ask for conditions that hold together or for some finite collections of properties where at least one of them must be true, system knowledge may be seen as a lattice structure. Those lattices of knowledge are introduced. The usage of lattice of knowledge perspective is exemplified within area of validating learning systems.