Skip to main content
Download PDF
- Main
Natural generation of admissible heuristics
Abstract
If a problem space can be represented by a relational production system, then many useful heuristics can be generated by appealing to appropriate abstraction spaces. Here we formalize the process by which heuristics can be generated. We show that these heuristics are admissible and monotonic. Finally we give several heuristics from the literature which could have been formed by the process, as well as some which cannot.