Applying a modified EQL optimization method to MRL rule-based programs

We modify our previously developed (Blaz Zupan et al., IEEE Trans. on Knowledge and Data Eng., April 1997) optimization method for EQL (EQuational rule-based Language) systems to optimize MRL (Macro Rule-based Language) systems. In particular, we show how the EQL optimization method can be applied to an MRL system after its corresponding state-space graphs have been constructed. Since the time and space complexity of a bidirectional search [O(b/sup d/2/)] is better than the breadth-first search's O(b/sup d/), we use bidirectional search and bidirectional breadth-first search strategies instead of the original bottom-up and breadth-first search strategies employed by Blaz Zupan et al. As in that paper, the resulting optimized MRL system (1) has a better response time in general because it requires fewer rule firings to reach the fixed point, (2) is stable because it has no cycles, and (3) has no redundant rules.