A fast approximation scheme for fractional covering problems with variable upper bounds

• Authors: Lisa Fleischer
• Venue: SODA '04 Proceedings of the fifteenth annual ACM-SIAM symposium on Discrete algorithms
• Year: 2004

We present the first combinatorial approximation scheme for mixed positive packing and covering linear programs that yields a pure approximation guarantee. Our algorithm returns solutions that simultaneously satisfy general positive covering constraints and packing constraints that are variable upper bounds. The returned solution has positive linear objective function value at most 1 + ε times the optimal value.Our approximation scheme is based on Lagrangian-relaxation methods. Previous such approximation schemes for mixed packing and covering problems does not simultaneously satisfy packing and covering constraints exactly. We show how to exactly satisfy general positive covering constraints simultaneously with variable upper bounds.A natural set of problems that our work addresses are linear programs for various network design problems: generalized Steiner network, vertex connectivity, directed connectivity, capacitated network design, group Steiner forest. These are all NP-hard problems for which there are approximation algorithms that round the solution to the corresponding linear program. Solving the linear program is often the computational bottleneck in these problems, and thus a fast approximation scheme for the LP relaxation means faster approximation algorithms.For the special case of survivable network design, we introduce a new modification of the push-relabel maximum flow algorithm that allows us to perform each iteration in amortized O(m + n log n) time, instead of one maximum flow per iteration that is implied by the straight forward adaptation of our general algorithm. (m is the number of edges and n is the number of vertices in the network.) In conjunction with an observation that reduces the number of iterations to {log n for f0} constraint matrices, the modification allows us to obtain an algorithm that is faster than existing exact or approximate algorithms by a factor of at least O(m) and by a factor of O(m log n) if the number of demand pairs is Ω(n).