Knowledge interpolation: A simple approach to rapid symbolic reasoning

The non-algorithmic nature of traditional knowledge-based (KB) reasoning ap proaches makes these techniques open-ended. As a result. KB techniques appe ar unsuitable for real-time applications as is evident from the small numbe r of appearances of KB techniques in real-time domains. The technique of kn owledge interpolation is a potential remedy for this shortcoming. Intuitive ly, the technique imitates the numerical-analysis technique of interpolatio n, derives solutions for an unknown problem from some already known values, and thereby avoids extensive searches of the knowledge base. An extra inhe rent advantage is that it gives the computation a more predictable algorith mic character. Hence not only can a computation's temporal requirements be estimated, but also requirements themselves may be reduced significantly. Effective application of this technique needs to answer two questions: what are the prerequisites for application of such techniques, and what are the possible ways of application. This paper studies both issues from an imple mentational point of view.