We introduce the $st$-cut version the Sparsest-Cut problem, where the goal is to find a cut of minimum sparsity among those separating two distinguished vertices $s,t\in V$. Clearly, this problem is at least as hard as the usual (non-$st$) version. Our main result is a polynomial-time algorithm for the product-demands setting, that produces a cut of sparsity $O(\sqrt{\OPT})$, where $\OPT$ denotes the optimum, and the total edge capacity and the total demand are assumed (by normalization) to be $1$. Our result generalizes the recent work of Trevisan [arXiv, 2013] for the non-$st$ version of the same problem (Sparsest-Cut with product demands), which in turn generalizes the bound achieved by the discrete Cheeger inequality, a cornerstone of Spectral Graph Theory that has numerous applications. Indeed, Cheeger's inequality handles graph conductance, the special case of product demands that are proportional to the vertex (capacitated) degrees. Along the way, we obtain an $O(\log n)$-approximation, where $n=\card{V}$, for the general-demands setting of Sparsest $st$-Cut.

Dependent rounding is a useful technique for optimization problems with hard budget constraints. This framework naturally leads to \emph{negative correlation} properties. However, what if an application naturally calls for dependent rounding on the one hand, and desires \emph{positive} correlation on the other? More generally, we develop algorithms that guarantee the known properties of dependent rounding, but also have nearly best-possible behavior -- near-independence, which generalizes positive correlation -- on ``small" subsets of the variables. The recent breakthrough of Li \& Svensson for the classical $k$-median problem has to handle positive correlation in certain dependent-rounding settings, and does so implicitly. We improve upon Li-Svensson's approximation ratio for $k$-median from $2.732 + \epsilon$ to $2.611 + \epsilon$ by developing an algorithm that improves upon various aspects of their work. Our dependent-rounding approach helps us improve the dependence of the runtime on the parameter $\epsilon$ from Li-Svensson's $N^{O(1/\epsilon^2)}$ to $N^{O((1/\epsilon) \log(1/\epsilon))}$.

In this paper, we study the uniform capacitated k-median problem. In the problem, we are given a set F of potential facility locations, a set C of clients, a metric d over F \cup C, an upper bound k on the number of facilities we can open and an upper bound u on the number of clients each facility can serve. We need to open a subset S \subseteq F of k facilities and connect clients in C to facilities in S so that each facility is connected by at most u clients. The goal is to minimize the total connection cost over all clients. Obtaining a constant approximation algorithm for this problem is a notorious open problem; most previous works gave constant approximations by either violating the capacity constraints or the cardinality constraint. Notably, all these algorithms are based on the natural LP-relaxation for the problem. The LP-relaxation has unbounded integrality gap, even when we are allowed to violate the capacity constraints or the cardinality constraint by a factor of 2-\eps. Our result is an \exp(O(1/\eps^2))-approximation algorithm for the problem that violates the cardinality constraint by a factor of 1+\eps. This is already beyond the capability of the natural LP relaxation, as it has unbounded integrality gap even if we are allowed to open (2-\eps)k facilities. Indeed, our result is based on a novel LP for this problem. We hope that this LP is the first step towards a constant approximation for capacitated k-median.

We consider the classic problem of envy-free division of a heterogeneous good ("cake") among several agents. It is known that, when the allotted pieces must be connected, the problem cannot be solved by a finite algorithm for 3 or more agents. Even when the pieces may be disconnected, no bounded-time algorithm is known for 5 or more agents. The impossibility result, however, assumes that the entire cake must be allocated. In this paper we replace the entire-allocation requirement with a weaker partial-proportionality requirement: the piece given to each agent must be worth for it at least a certain positive fraction of the entire cake value. We prove that this version of the problem is solvable in bounded time even when the pieces must be connected. We present bounded-time envy-free cake-cutting algorithms for: (1) giving each of $n$ agents a connected piece with a positive value; (2) giving each of 3 agents a connected piece worth at least 1/3; (3) giving each of 4 agents a connected piece worth at least 1/7; (4) giving each of 4 agents a disconnected piece worth at least 1/4; (5) giving each of $n$ agents a disconnected piece worth at least $(1-\epsilon)/n$ for any positive $\epsilon$.

This paper studies the set cover problem under the semi-streaming model. The underlying set system is formalized in terms of a hypergraph $G = (V, E)$ whose edges arrive one-by-one and the goal is to construct an edge cover $F \subseteq E$ with the objective of minimizing the cardinality (cost in the weighted case) of $F$. We consider a parameterized relaxation of this problem, where given some $0 \leq \epsilon < 1$, the goal is to construct an edge $(1 - \epsilon)$-cover, namely, a subset of edges incident to all but an $\epsilon$-fraction of the vertices (or their benefit in the weighted case). The key limitation imposed on the algorithm is that its space is limited to (poly)logarithmically many bits per vertex. Our main result is an asymptotically tight trade-off between $\epsilon$ and the approximation ratio: We design a semi-streaming algorithm that on input hypergraph $G$, constructs a succinct data structure $\mathcal{D}$ such that for every $0 \leq \epsilon < 1$, an edge $(1 - \epsilon)$-cover that approximates the optimal edge \mbox{($1$-)cover} within a factor of $f(\epsilon, n)$ can be extracted from $\mathcal{D}$ (efficiently and with no additional space requirements), where \[ f(\epsilon, n) = \left\{ \begin{array}{ll} O (1 / \epsilon), & \text{if } \epsilon > 1 / \sqrt{n} \\ O (\sqrt{n}), & \text{otherwise} \end{array} \right. \, . \] In particular for the traditional set cover problem we obtain an $O(\sqrt{n})$-approximation. This algorithm is proved to be best possible by establishing a family (parameterized by $\epsilon$) of matching lower bounds.

We consider the problem of, given an array $A[1,n]$ of elements with a total order, building a data structure that solves two queries: $(a)$ selection queries receive a range $[i,j]$ and an integer $k$ and return the position of the $k$th largest element in $A[i,j]$; $(b)$ top-$k$ queries receive $[i,j]$ and $k$ and return the positions of the $k$ largest elements in $A[i,j]$. These problems can be solved in optimal time, $O(1+\log k/\log\log n)$ and $O(k)$ respectively, using linear-space data structures. We present the first study of {\em encoding} data structures for this problem, which do not access $A$ at query time. These save storage space in applications where the values of $A$ themselves are not of interest. We first show that any encoding answering such queries requires $n\lg k - O(n+k \lg k)$ bits of space. Then we design encodings using $O(n\log k)$ bits, that is, asymptotically optimal up to constant factors, that answer selection and top-$k$ queries in optimal time.

#### Algorithms for Hub Label Optimization

Maxim Babenko(); Andrew Goldberg(); Prof. Anupam Gupta(Carnegie Mellon University); Dr. Viswanath Nagarajan(University of Michigan)Even though local search heuristics are the method of choice in practice for many well-studied optimization problems, most of them behave poorly in the worst case. This is in particular the case for the Maximum-Cut Problem, for which local search can take an exponential number of steps to terminate and the problem of computing a local optimum is PLS-complete. To narrow the gap between theory and practice, we study local search for the Maximum-Cut Problem in the framework of smoothed analysis in which inputs are subject to a small amount of random noise. We show that the smoothed number of iterations is quasi-polynomial, i.e., it is bounded from above by a polynomial in n^log(n) and ¦ where n denotes the number of nodes and ¦ denotes the perturbation parameter. This shows that worst-case instances are fragile and it is a first step in explaining why they are rarely observed in practice.

We propose a new approach to competitive analysis in online scheduling by introducing the novel concept of competitive-ratio approximation schemes. Such a scheme algorithmically constructs an online algorithm with a competitive ratio arbitrarily close to the best possible competitive ratio for any online algorithm. We study the problem of scheduling jobs online to minimize the weighted sum of completion times on parallel, related, and unrelated machines, and we derive both deterministic and randomized algorithms which are almost best possible among all online algorithms of the respective settings. We also generalize our techniques to arbitrary monomial cost functions and apply them to the makespan objective. Our method relies on an abstract characterization of online algorithms combined with various simplifications and transformations. We also contribute algorithmic means to compute the actual value of the best possible competitive ratio up to an arbitrary accuracy. This strongly contrasts nearly all previous manually obtained competitiveness results and, most importantly, it reduces the search for the optimal competitive ratio to a question that a computer can answer. We believe that our concept can also be applied to many other problems and yields a new perspective on online algorithms in general.

We study the Minimum Latency Submodular Cover problem (MLSC), which consists of a metric $(V,d)$ with source $r\in V$ and $m$ monotone submodular functions $f_1, f_2, ..., f_m: 2^V \rightarrow [0,1]$. The goal is to find a path originating at $r$ that minimizes the total cover time of all functions. This generalizes well-studied problems, such as Submodular Ranking [AzarG11] and Group Steiner Tree [GargKR00]. We give a polynomial time $O(\log \frac{1}{\eps} \cdot \log^{2+\delta} |V|)$-approximation algorithm for MLSC, where $\epsilon>0$ is the smallest non-zero marginal increase of any $\{f_i\}_{i=1}^m$ and $\delta>0$ is any constant. We also consider the Latency Covering Steiner Tree problem (LCST), which is the special case of \mlsc where the $f_i$s are multi-coverage functions. This is a common generalization of the Latency Group Steiner Tree [GuptaNR10, ChakrabartyS11] and Generalized Min-sum Set Cover [AzarGY09, BansalGK10] problems. We obtain an $O(\log^2|V|)$-approximation algorithm for LCST. Finally we study a natural stochastic extension of the Submodular Ranking problem, and obtain an adaptive algorithm with an $O(\log 1/ \eps)$ approximation ratio, which is best possible. This result also generalizes some previously studied stochastic optimization problems, such as Stochastic Set Cover [GoemansV06] and Shared Filter Evaluation [MunagalaSW07,LiuPRY08].

Several simple, classical, little-known algorithms in the statistical literature for generating random permutations by coin-tossing are examined, analyzed and implemented. These algorithms are either asymptotically optimal or close to being so in terms of the expected number of times the random bits are generated. In addition to asymptotic approximations to the expected complexity, we also clarify the corresponding variances, as well as the asymptotic distributions. A brief comparative discussion with numerical computations in a multicore system is also given.

The dynamic facility location problem is a generalization of the classic facility location problem proposed by Eisenstat, Mathieu, and Schabanel to model the dynamics of evolving social/infrastructure networks. The generalization lies in that the distance metric between clients and facilities changes over time. This leads to a trade-off between optimizing the classic objective function and the "stability" of the solution: there is a switching cost charged every time a client changes the facility to which it is connected. While the standard linear program (LP) relaxation for the classic problem naturally extends to this problem, traditional LP-rounding techniques do not, as they are often sensitive to small changes in the metric resulting in frequent switches. We present a new LP-rounding algorithm for facility location problems, which yields the first constant approximation algorithm for the dynamic facility location problem. Our algorithm installs competing exponential clocks on the clients and facilities, and connect every client by the path that repeatedly follows the smallest clock in the neighborhood. The use of exponential clocks gives rise to several properties that distinguish our approach from previous LP-roundings for facility location problems. In particular, we use no clustering and we allow clients to connect through paths of arbitrary lengths. In fact, the clustering-free nature of our algorithm is crucial for applying our LP-rounding approach to the dynamic problem.