Approximation Algorithms for Path TSP, ATSP, and TAP via Relaxations

Linear programming (LP) relaxations provide a powerful technique to design approximation algorithms for combinatorial optimization problems. In the first part of the thesis, we study the metric s-t path Traveling Salesman Problem (TSP) via LP relaxations. We first consider the s-t path graph-TSP, a critical special case of the metric s-t path TSP. We design a new simple LP-based algorithm for the s-t path graph-TSP that achieves the best known approximation factor of 1.5. Then, we turn our attention to the general metric s-t path TSP. [An, Kleinberg, and Shmoys, STOC 2012] improved on the long standing 5 3 -approximation factor and presented an algorithm that achieves an approximation factor of 1+ √ 5 2 ≈ 1.61803. Later, [Sebő, IPCO 2013] further improved the approximation factor to 8 5 . We present a simple, self-contained analysis that unifies both results. Additionally, we compare two different LP relaxations of the s-t path TSP, namely the path version of the Held-Karp LP relaxation for TSP and a weaker LP relaxation, and we show that both LPs have the same (fractional) optimal value. Also, we show that the minimum cost of integral solutions of the two LPs are within a factor of 3 2 of each other. Furthermore, we prove that a half-integral solution of the stronger LP relaxation of cost c can be rounded to an integral solution of cost at most 3 2 c. Finally, we give an instance that presents obstructions to two natural methods that aim for an approximation factor of 3 2 . The Sherali-Adams (SA) system and the Lasserre (Las) system are two popular Lift-andProject systems that tighten a given LP relaxation in a systematic way. In the second part of the thesis, we study the Asymmetric Traveling Salesman Problem (ATSP) and unweighted Tree Augmentation Problem, respectively, in the framework of the SA system and the Las system. For ATSP, our focus is on negative results. For any fixed integer t ≥ 0 and small , 0 < 1, we prove that the integrality ratio for level t of the SA system starting with the standard LP relaxation of ATSP is at least 1 + 1− 2t+3 . For a further relaxation of ATSP called the balanced LP relaxation, we obtain an integrality ratio lower bound of 1 + 1− t+1 for level t of the SA system. Also, our results for the standard LP relaxation extend to the path version of ATSP. For the unweighted Tree Augmentation Problem, our focus is on positive results. We study this problem via the Las system. We prove an upper bound of (1.5 + ) on the integrality ratio of a semidefinite programming (SDP) relaxation, where > 0 can be any small constant, by analyzing a combinatorial algorithm. This SDP relaxation is derived by applying the Las system to an initial LP relaxation. We generalize the combinatorial analysis of integral solutions from the previous literature to fractional solutions by identifying some properties of fractional solutions of the Las system via the decomposition result of [Karlin, Mathieu, and Nguyen, IPCO 2011].