Answer-Set Programming (ASP) is an expressive rule-based knowledge-representation formalism supported by efficient solver technology. Traditional evaluation of answer-set programs takes place in two phases: grounding and solving. Grounding incurs an up-to exponential increase in space, termed the grounding bottleneck of ASP, which is often encountered in practice. Lazy grounding avoids this bottleneck but is restricted to normal rules, significantly limiting the expressive power of this approach. We propose a framework to handle aggregates by normalizing them on demand during the lazy grounding process; we call this approach lazy normalization. It is feasible for different types of aggregates and can bring about up-to exponential gains in space and time.
|Number of pages||2|
|Publication status||Published - 2018|
- Knowledge Representation, Lazy Grounding, Normalization, Nonmonotonic Reasoning, Answer-Set Programming