arXiv subjects

Konstantinos Georgiou

Publications and source records attributed to Konstantinos Georgiou.

41 records · Page 3Linked to original sources

arXiv2013

Network bargaining with general capacities

We study balanced solutions for network bargaining games with general capacities, where agents can participate in a fixed but arbitrary number of contracts. We provide the first polynomial time algorithm for computing balanced solutions for these games. In addition, we prove that an instance has a balanced solution if and only if it has a stable one. Our methods use a new idea of reducing an instance with general capacities to a network bargaining game with unit capacities defined on an auxiliary graph. This represents a departure from previous approaches, which rely on computing an allocation in the intersection of the core and prekernel of a corresponding cooperative game, and then proving that the solution corresponding to this allocation is balanced. In fact, we show that such cooperative game methods do not extend to general capacity games, since contrary to the case of unit capacities, there exist allocations in the intersection of the core and prekernel with no corresponding balanced solution. Finally, we identify two sufficient conditions under which the set of balanced solutions corresponds to the intersection of the core and prekernel, thereby extending the class of games for which this result was previously known.

cs.GT↗

arXiv2013

The Beachcombers' Problem: Walking and Searching with Mobile Robots

We introduce and study a new problem concerning the exploration of a geometric domain by mobile robots. Consider a line segment $[0,I]$ and a set of $n$ mobile robots $r_1,r_2,..., r_n$ placed at one of its endpoints. Each robot has a {\em searching speed} $s_i$ and a {\em walking speed} $w_i$, where $s_i <w_i$. We assume that each robot is aware of the number of robots of the collection and their corresponding speeds. At each time moment a robot $r_i$ either walks along a portion of the segment not exceeding its walking speed $w_i$ or searches a portion of the segment with the speed not exceeding $s_i$. A search of segment $[0,I]$ is completed at the time when each of its points have been searched by at least one of the $n$ robots. We want to develop {\em mobility schedules} (algorithms) for the robots which complete the search of the segment as fast as possible. More exactly we want to maximize the {\em speed} of the mobility schedule (equal to the ratio of the segment length versus the time of the completion of the schedule). We analyze first the offline scenario when the robots know the length of the segment that is to be searched. We give an algorithm producing a mobility schedule for arbitrary walking and searching speeds and prove its optimality. Then we propose an online algorithm, when the robots do not know in advance the actual length of the segment to be searched. The speed $S$ of such algorithm is defined as $S = \inf_{I_L} S(I_L)$ where $S(I_L)$ denotes the speed of searching of segment $I_L=[0,L]$. We prove that the proposed online algorithm is 2-competitive. The competitive ratio is shown to be better in the case when the robots' walking speeds are all the same.

cs.DS↗

arXiv2012

Understanding Set Cover: Sub-exponential Time Approximations and Lift-and-Project Methods

Recently, Cygan, Kowalik, and Wykurz [IPL 2009] gave sub-exponential-time approximation algorithms for the Set-Cover problem with approximation ratios better than ln(n). In light of this result, it is natural to ask whether such improvements can be achieved using lift-and-project methods. We present a simpler combinatorial algorithm which has nearly the same time-approximation tradeoff as the algorithm of Cygan et al., and which lends itself naturally to a lift-and-project based approach. At a high level, our approach is similar to the recent work of Karlin, Mathieu, and Nguyen [IPCO 2011], who examined a known PTAS for Knapsack (similar to our combinatorial Set-Cover algorithm) and its connection to hierarchies of LP and SDP relaxations for Knapsack. For Set-Cover, we show that, indeed, using the trick of "lifting the objective function", we can match the performance of our combinatorial algorithm using the LP hierarchy of Lovasz and Schrijver. We also show that this trick is essential: even in the stronger LP hierarchy of Sherali and Adams, the integrality gap remains at least (1-eps) ln(n) at level Omega(n) (when the objective function is not lifted). As shown by Aleknovich, Arora, and Tourlakis [STOC 2005], Set-Cover relaxations stemming from SDP hierarchies (specifically, LS+) have similarly large integrality gaps. This stands in contrast to Knapsack, where Karlin et al. showed that the (much stronger) Lasserre SDP hierarchy reduces the integrality gap to (1+eps) at level O(1). For completeness, we show that LS+ also reduces the integrality gap for Knapsack to (1+eps). This result may be of independent interest, as our LS+ based rounding and analysis are rather different from those of Karlin et al., and to the best of our knowledge this is the first explicit demonstration of such a reduction in the integrality gap of LS+ relaxations after few rounds.

math.CO↗

arXiv2012

Better Balance by Being Biased: A 0.8776-Approximation for Max Bisection

Recently Raghavendra and Tan (SODA 2012) gave a 0.85-approximation algorithm for the Max Bisection problem. We improve their algorithm to a 0.8776-approximation. As Max Bisection is hard to approximate within $α_{GW} + ε\approx 0.8786$ under the Unique Games Conjecture (UGC), our algorithm is nearly optimal. We conjecture that Max Bisection is approximable within $α_{GW}-ε$, i.e., the bisection constraint (essentially) does not make Max Cut harder. We also obtain an optimal algorithm (assuming the UGC) for the analogous variant of Max 2-Sat. Our approximation ratio for this problem exactly matches the optimal approximation ratio for Max 2-Sat, i.e., $α_{LLZ} + ε\approx 0.9401$, showing that the bisection constraint does not make Max 2-Sat harder. This improves on a 0.93-approximation for this problem due to Raghavendra and Tan.

cs.DS↗

arXiv2012

Efficient Algorithms for Solving Hypergraphic Steiner Tree Relaxations in Quasi-Bipartite Instances

We consider the Steiner tree problem in quasi-bipartite graphs, where no two Steiner vertices are connected by an edge. For this class of instances, we present an efficient algorithm to exactly solve the so called directed component relaxation (DCR), a specific form of hypergraphic LP relaxation that was instrumental in the recent break-through result by Byrka et al. [BGRS10] (STOC 2010). Our algorithm hinges on an efficiently computable map from extreme points of the bidirected cut relaxation to feasible solutions of (DCR). As a consequence, together with [BGRS10] we immediately obtain an efficient 73/60-approximation for quasi-bipartite Steiner tree instances. We also present a particularly simple (BCR)-based random sampling algorithm that achieves a performance guarantee slightly better than 77/60.

cs.DM↗