MS 101 Algorithms University of Delhi德里大学101女士.pptVIP

MCS 312: NP Completeness and Approximation algorthms Instructor Neelima Gupta ngupta@cs.du.ac.in Table of Contents O(log n) factor Approximation for Set Cover using Greedy Approach Reference: Vazirani Set Cover Problem Given: A Universe U of n elements A collection of subsets of U, S = {S1, S2…..Sk} A cost function c:S→Q+ To Find: A minimum cost subcollection of S that covers all elements of U. Some Notations and Definitions C : set of elements already covered at the beginning of an iteration. Cost Effectiveness of S: Average cost at which it covers new elements, i.e. c(S) / |S-C|. Price of an element: Average cost at which it is covered, i.e. price(e) = c(S) / |S-C|. Greedy Set Cover Algorithm C←Ф 2. While C ≠ U do Find the most cost-effective set in the current iteration, say S. Let α= ,i.e., the cost effectiveness of S Pick S, and for each e in S-C, price(e)= α C ← C S Output the picked sets. Elements of U are then numbered in the order in which they were covered by the algorithm. Let this numbering be: e1,e2,……..,en Greedy Set Cover Algorithm Claim : For any k {1,2,….,n}, price(ek) Proof Observations At any point of time, some of the sets of optimal solution may be present in our solution, and some may not be. In the worst case, our solution would currently not be containing any set from the optimal solution. In this case, the leftover sets of the optimal solution (actually all sets belonging to optimal), would therefore cover the remaining elements at a cost no more than OPT. Claim: at any iteration, there exists an element e which can be covered by a set at cost = OPT/ |S-C| Proof: Suppose if possible all the remaining elements need to pay a cost OPT/ |S-C| Then, this means that the remaining elements cannot be covered at cost = |S-C| * OPT/ |S-C| = OPT But this