Prolog with best first search

Prolog is a logic programming language with an engine that can search a knowledge base with respect to a query automatically. To satisfy a query of a knowledge base, the Prolog engine always try to use the top-most clause and search the left-most sub query, which is fundamentally a depth first search (DFS). While generally usable, DFS is less effective in many circumstances, and may even fail to discover solutions in some extreme cases. This paper presents an alternative best first search (BFS) for the Prolog engine. Our BFD search strategy employes a heuristics function based on an extended measurement of UCB1, which is a machine learning algorithm for a Markovian Decision Process (MDP). This heuristics function can balance exploitation and exploration during the search process of the Prolog engine, adjust its search path dynamically for the best chance to satisfy its query based on its accumulated knowledge of the knowledge base. Compared with DFS Prolog, BFD Prolog not only improves its search efficiency, but also guarantees theoretically to discover its solutions if they exist. Unlike DFS Prolog, BFS Prolog is completely declarative as knowledge orders are irrelevant.