组合算法基础教学课件.ppt

1 Finding k-thMinimum Another searching problem is to find the k-th smallest element in a sequence S of n element. One obvious method is to sort the sequence into non decreasing order (by Merge-Sort Algorithm) and then locate the k-th element. By Theorem 3 this method would require O(nlogn) comparisons. By a careful application of the divide-and-conquer strategy, we can find the k-th smallest element in O(n) comparisons. The basic idea is to partition the given sequence into three subsequences S1, S2, S3 such that S1 contains all elements less than m, S2 all elements equal to m, and S3 all elements greater than m. By counting the number of elements in S1 and in S2 we can determine in which of S1, S2, or S3 the k-th smallest element lies. In such a way we can reduce the problem to a smaller subproblem. In order to obtain a linear-time algorithm we must be able to find a partition element in linear time such that the sizes of the subsequences S1 and S3 are each no more than a fixed fraction of the size of S. The crux is in how the partition element m is chosen. The sequence S is partitioned into subsequences of five elements each. Each of them is sorted and from which the median is selected, medians of all subsequences form a sequence M, which contains onlyn/5elements. We can find the median of M five times faster than that of a sequence of n elements. 2 Dynamic Programming The divide-and-conquer approach partitions the problem into independent sub-problems, solves the sub-problems recursively, and then combine their solutions to solve the original problem. In contrast, the dynamic programming approach is applicable when the sub-problems are not independent, i.e., when subproblems share subsubproblems. It usually consists of the following steps: ? characterize the structure of an optimal solution, ? recursively define the value of an optimal solution, ? compute the value of an optimal solution, ? construct an optimal solution from comput