2015AI课件lec4启发式搜索2013章节.pptVIP

Admissibility of A* algorithms Breadth-first search may be characterized as an A* algorithm in which f(n)=g(n)+0. the set of nodes considered by an A* algorithm is a subset of the states examined in breadth-first search. Eg1: Eight-Puzzle Start Goal g(n) = the cost of the lowest-cost path found (from n0 to n). h(n) =number of tiles out of place (compared with goal) ≤ h(n) The intensity of heuristic information Strong: Can reduce the quantity of search process, but cannot guarantee of finding the optimal solution. Weak: Generally, lead to great search process, even become uninformed search, but have the chance to find the optimal solution. The intensity of heuristic information 0 Computation cost Quantity of information Cost of rule applying Cost of control strategy Cost of solving the problem Optimal solution Completely informtion = + Eg1: Eight-Puzzle Start Goal g(n) = the cost of the lowest-cost path found (from n0 to n). h(n) =number of tiles out of place (compared with goal) ≤ h(n) 9 Other heuristic search Other heuristic search Iterative-Deepening A* (IDA*) Recursive Best-First Search Bidirectional Search Iterative-Deepening A* (IDA*)迭代加深的A* 进行一系列带有“截止代价”的深度优先搜索，搜索时不生成超过截止代价的后继节点。初始截止代价定义为： 如果找到目标节点就结束搜索，否则增大截断值，开始另一次深度优先搜索。上一次深度优先搜索中遍历过的（但没有扩展）的节点的最小 值作为下一次深度优先搜索的新的截断值。 Eg:罗马尼亚寻径 Optimal complete Iterative-Deepening A* (IDA*) Recursive Best-First Search递归最优搜索 Backing up value When a node n is expanded, computes values of successors of n and recomputes values of n and all of n’s ancestors Process of backing up Backed-up value, , of node m with successors Description One of successors of node n has the smallest over all OPEN nodes, it is expanded in turn, and so on. When other OPEN node, n’, (not a successor of n) has the lowest value of backtracks to the lowest common ancestor, node k kn: successor of node k on the path to n RBFS removes the subtree