Exactly how good are heuristics? Toward a realistic predictive theory of best-first search

Open in new window