摘要: We introduce a model for predicting the performance of IDA* using pattern database heuristics, as function branching factor problem, solution depth, and size databases. While it is known that larger database, more efficient search, we provide quantitative analysis this relationship. In particular, show single goal state, number nodes expanded by fraction (logb s + 1)/s brute-force where b factor, database. also taking maximum at least two databases, node expansions decreases linearly with compared to search. compare our theoretical predictions empirical data on Rubik's Cube. Our conservative, overestimates actual expansions.