Results 11  20
of
317
Measured descent: A new embedding method for finite metrics
 In Proc. 45th FOCS
, 2004
"... We devise a new embedding technique, which we call measured descent, based on decomposing a metric space locally, at varying speeds, according to the density of some probability measure. This provides a refined and unified framework for the two primary methods of constructing Fréchet embeddings for ..."
Abstract

Cited by 98 (32 self)
 Add to MetaCart
(Show Context)
We devise a new embedding technique, which we call measured descent, based on decomposing a metric space locally, at varying speeds, according to the density of some probability measure. This provides a refined and unified framework for the two primary methods of constructing Fréchet embeddings for finite metrics, due to [Bourgain, 1985] and [Rao, 1999]. We prove that any npoint metric space (X, d) embeds in Hilbert space with distortion O ( √ αX · log n), where αX is a geometric estimate on the decomposability of X. As an immediate corollary, we obtain an O ( √ (log λX)log n) distortion embedding, where λX is the doubling constant of X. Since λX ≤ n, this result recovers Bourgain’s theorem, but when the metric X is, in a sense, “lowdimensional, ” improved bounds are achieved. Our embeddings are volumerespecting for subsets of arbitrary size. One consequence is the existence of (k, O(log n)) volumerespecting embeddings for all 1 ≤ k ≤ n, which is the best possible, and answers positively a question posed by U. Feige. Our techniques are also used to answer positively a question of Y. Rabinovich, showing that any weighted npoint planar graph O(log n) embeds in ℓ∞ with O(1) distortion. The O(log n) bound on the dimension is optimal, and improves upon the previously known bound of O((log n) 2). 1
A Combinatorial, PrimalDual approach to Semidefinite Programs
"... Semidefinite programs (SDP) have been used in many recent approximation algorithms. We develop a general primaldual approach to solve SDPs using a generalization of the wellknown multiplicative weights update rule to symmetric matrices. For a number of problems, such as Sparsest Cut and Balanced ..."
Abstract

Cited by 95 (12 self)
 Add to MetaCart
(Show Context)
Semidefinite programs (SDP) have been used in many recent approximation algorithms. We develop a general primaldual approach to solve SDPs using a generalization of the wellknown multiplicative weights update rule to symmetric matrices. For a number of problems, such as Sparsest Cut and Balanced Separator in undirected and directed weighted graphs, and the Min UnCut problem, this yields combinatorial approximation algorithms that are significantly more efficient than interior point methods. The design of our primaldual algorithms is guided by a robust analysis of rounding algorithms used to obtain integer solutions from fractional ones.
The little engine(s) that could: scaling online social networks
 in ACM SIGCOMM Conference, 2010
"... The difficulty of scaling Online Social Networks (OSNs) has introduced new system design challenges that has often caused costly rearchitecting for services like Twitter and Facebook. The complexity of interconnection of users in social networks has introduced new scalability challenges. Convention ..."
Abstract

Cited by 67 (5 self)
 Add to MetaCart
(Show Context)
The difficulty of scaling Online Social Networks (OSNs) has introduced new system design challenges that has often caused costly rearchitecting for services like Twitter and Facebook. The complexity of interconnection of users in social networks has introduced new scalability challenges. Conventional vertical scaling by resorting to full replication can be a costly proposition. Horizontal scaling by partitioning and distributing data among multiples servers – e.g. using DHTs – can lead to costly interserver communication. We design, implement, and evaluate SPAR, a social partitioning and replication middleware that transparently leverages the social graph structure to achieve data locality while minimizing replication. SPAR guarantees that for all users in an OSN, their direct neighbor’s data is colocated in the same server. The gains from this approach are multifold: application developers can assume local semantics, i.e., develop as they would for a single server; scalability is achieved by adding commodity servers with low memory and network I/O requirements; and redundancy is achieved at a fraction of the cost. We detail our system design and an evaluation based on datasets from Twitter, Orkut, and Facebook, with a working implementation. We show that SPAR incurs minimum overhead, and can help a wellknown opensource Twitter clone reach Twitter’s scale without changing a line of its application logic and achieves higher throughput than Cassandra, Facebook’s DHT based keyvalue store database.
Inoculation Strategies for Victims of Viruses and the SumofSquares Partition Problem
 PROCEEDINGS OF THE 16TH ANNUAL ACMSIAM SYMPOSIUM ON DISCRETE ALGORITHMS
, 2005
"... We propose a simple game for modeling containment of the spread of viruses in a graph of n nodes. Each node must choose to either install antivirus software at some known cost C, or risk infection and a loss L if a virus that starts at a random initial point in the graph can reach it without being ..."
Abstract

Cited by 67 (2 self)
 Add to MetaCart
We propose a simple game for modeling containment of the spread of viruses in a graph of n nodes. Each node must choose to either install antivirus software at some known cost C, or risk infection and a loss L if a virus that starts at a random initial point in the graph can reach it without being stopped by some intermediate node. The goal of individual nodes is to minimize their individual expected cost. We prove many game theoretic properties of the model, including an easily applied characterization of Nash equilibria, culminating in our showing that allowing selfish users to choose Nash equilibrium strategies is highly undesirable, because the price of anarchy is an unacceptable Θ(n) in the worst case. This shows in particular that a centralized solution can give a much better total cost than an equilibrium solution. Though it is NPhard to compute such a social optimum, we show that the problem can be reduced to a previously unconsidered combinatorial problem that we call the sumofsquares partition problem. Using a greedy algorithm based on sparse cuts, we show that this problem can be approximated to within a factor of O(log² n), giving the same approximation ratio for the inoculation game.
Balanced graph partitioning
 In 16th Annual ACM Symposium on Parallelism in Algorithms and Architectures
, 2004
"... We consider the problem of partitioning a graph into k components of roughly equal size while minimizing the capacity of the edges between different components of the cut. In particular we require that for a parameter ν ≥ 1, no component contains more than ν · n k of the graph vertices. For k = 2 an ..."
Abstract

Cited by 67 (0 self)
 Add to MetaCart
We consider the problem of partitioning a graph into k components of roughly equal size while minimizing the capacity of the edges between different components of the cut. In particular we require that for a parameter ν ≥ 1, no component contains more than ν · n k of the graph vertices. For k = 2 and ν = 1 this problem is equivalent to the well known Minimum Bisection Problem for which an approximation algorithm with a polylogarithmic approximation guarantee has been presented in [FK02]. For arbitrary k and ν ≥ 2 a bicriteria approximation ratio of O(logn) was obtained by [ENRS99] using the spreading metrics technique. We present a bicriteria approximation algorithm that for any constant ν> 1 runs in polynomial time and guarantees an approximation ratio of O(log1.5 n) (for a precise statement of the main result see Theorem 6). For ν = 1 and k ≥ 3 we show that no polynomial time approximation algorithm can guarantee a finite approximation ratio unless P = NP. 1
A better approximation ratio for the vertex cover problem
, 2005
"... We reduce the approximation factor for Vertex Cover to 2 − Θ ( 1 √ log n) (instead of the previous log log n 2 − Θ ( log n), obtained by BarYehuda and Even [2], and by Monien and Speckenmeyer [10]). The improvement of the vanishing factor comes as an application of the recent results of Arora, Rao, ..."
Abstract

Cited by 65 (0 self)
 Add to MetaCart
(Show Context)
We reduce the approximation factor for Vertex Cover to 2 − Θ ( 1 √ log n) (instead of the previous log log n 2 − Θ ( log n), obtained by BarYehuda and Even [2], and by Monien and Speckenmeyer [10]). The improvement of the vanishing factor comes as an application of the recent results of Arora, Rao, and Vazirani [1] that improved the approximation factor of the sparsest cut and balanced cut problems. In particular, we use the existence of two big and wellseparated sets of nodes in the solution of the semidefinite relaxation for balanced cut, proven in [1]. We observe that a solution of the semidefinite relaxation for vertex cover, when strengthened with the triangle inequalities, can be transformed into a solution of a balanced cut problem, and therefore the existence of big wellseparated sets in the sense of [1] translates into the existence of a big independent set. 1
Optimal Hierarchical Decompositions for Congestion Minimization in Networks
, 2008
"... Hierarchical graph decompositions play an important role in the design of approximation and online algorithms for graph problems. This is mainly due to the fact that the results concerning the approximation of metric spaces by tree metrics (e.g. [10, 11, 14, 16]) depend on hierarchical graph decompo ..."
Abstract

Cited by 63 (2 self)
 Add to MetaCart
Hierarchical graph decompositions play an important role in the design of approximation and online algorithms for graph problems. This is mainly due to the fact that the results concerning the approximation of metric spaces by tree metrics (e.g. [10, 11, 14, 16]) depend on hierarchical graph decompositions. In this line of work a probability distribution over tree graphs is constructed from a given input graph, in such a way that the tree distances closely resemble the distances in the original graph. This allows it, to solve many problems with a distancebased cost function on trees, and then transfer the tree solution to general undirected graphs with only a logarithmic loss in the performance guarantee. The results about oblivious routing [30, 22] in general undirected graphs are based on hierarchical decompositions of a different type in the sense that they are aiming to approximate the bottlenecks in the network (instead of the pointtopoint distances). We call such decompositions cutbased decompositions. It has been shown that they also can be used to design approximation and online algorithms for a wide variety of different problems, but at the current state of the art the performance guarantee goes down by an O(log 2 n log log n)factor when making the transition from tree networks to general graphs. In this paper we show how to construct cutbased decompositions that only result in a logarithmic loss in performance, which is asymptotically optimal. Remarkably, one major ingredient of our proof is a distancebased decomposition scheme due to Fakcharoenphol, Rao and Talwar [16]. This shows an interesting relationship between these seemingly different decomposition techniques. The main applications of the new decomposition are an optimal O(log n)competitive algorithm for oblivious routing in general undirected graphs, and an O(log n)approximation for Minimum Bisection, which improves the O(log 1.5 n) approximation
A local clustering algorithm for massive graphs and its application to nearlylinear time graph partitioning
, 2013
"... We study the design of local algorithms for massive graphs. A local graph algorithm is one that finds a solution containing or near a given vertex without looking at the whole graph. We present a local clustering algorithm. Our algorithm finds a good cluster—a subset of vertices whose internal conn ..."
Abstract

Cited by 58 (8 self)
 Add to MetaCart
(Show Context)
We study the design of local algorithms for massive graphs. A local graph algorithm is one that finds a solution containing or near a given vertex without looking at the whole graph. We present a local clustering algorithm. Our algorithm finds a good cluster—a subset of vertices whose internal connections are significantly richer than its external connections—near a given vertex. The running time of our algorithm, when it finds a nonempty local cluster, is nearly linear in the size of the cluster it outputs. The running time of our algorithm also depends polylogarithmically on the size of the graph and polynomially on the conductance of the cluster it produces. Our clustering algorithm could be a useful primitive for handling massive graphs, such as social networks and webgraphs. As an application of this clustering algorithm, we present a partitioning algorithm that finds an approximate sparsest cut with nearly optimal balance. Our algorithm takes time nearly linear in the number edges of the graph. Using the partitioning algorithm of this paper, we have designed a nearly linear time algorithm for constructing spectral sparsifiers of graphs, which we in turn use in a nearly linear time algorithm for solving linear systems in symmetric, diagonally dominant matrices. The linear system solver also leads to a nearly linear time algorithm for approximating the secondsmallest eigenvalue and corresponding eigenvector of the Laplacian matrix of a graph. These other results are presented in two companion papers.
Approximate Clustering without the Approximation
"... Approximation algorithms for clustering points in metric spaces is a flourishing area of research, with much research effort spent on getting a better understanding of the approximation guarantees possible for many objective functions such as kmedian, kmeans, and minsum clustering. This quest for ..."
Abstract

Cited by 55 (19 self)
 Add to MetaCart
(Show Context)
Approximation algorithms for clustering points in metric spaces is a flourishing area of research, with much research effort spent on getting a better understanding of the approximation guarantees possible for many objective functions such as kmedian, kmeans, and minsum clustering. This quest for better approximation algorithms is further fueled by the implicit hope that these better approximations also give us more accurate clusterings. E.g., for many problems such as clustering proteins by function, or clustering images by subject, there is some unknown “correct” target clustering and the implicit hope is that approximately optimizing these objective functions will in fact produce a clustering that is close (in symmetric difference) to the truth. In this paper, we show that if we make this implicit assumption explicit—that is, if we assume that any capproximation to the given clustering objective F is ǫclose to the target—then we can produce clusterings that are O(ǫ)close to the target, even for values c for which obtaining a capproximation is NPhard. In particular, for kmedian and kmeans objectives, we show that we can achieve this guarantee for any constant c> 1, and for minsum objective we can do this for any constant c> 2. Our results also highlight a somewhat surprising conceptual difference between assuming that the optimal solution to, say, the kmedian objective is ǫclose to the target, and assuming that any approximately optimal solution is ǫclose to the target, even for approximation factor say c = 1.01. In the former case, the problem of finding a solution that is O(ǫ)close to the target remains computationally hard, and yet for the latter we have an efficient algorithm.
Embeddings of negativetype metrics and an improved approximation to generalized sparsest cut
, 2007
"... In this paper, we study metrics of negative type, which are metrics (V, d) such that √ d is an Euclidean metric; these metrics are thus also known as “ℓ2squared” metrics. We show how to embed npoint negativetype metrics into Euclidean space ℓ2 with distortion D = O(log 3/4 n). This embedding resu ..."
Abstract

Cited by 52 (0 self)
 Add to MetaCart
(Show Context)
In this paper, we study metrics of negative type, which are metrics (V, d) such that √ d is an Euclidean metric; these metrics are thus also known as “ℓ2squared” metrics. We show how to embed npoint negativetype metrics into Euclidean space ℓ2 with distortion D = O(log 3/4 n). This embedding result, in turn, implies an O(log 3/4 k)approximation algorithm for the Sparsest Cut problem with nonuniform demands. Another corollary we obtain is that npoint subsets of ℓ1 embed into ℓ2 with distortion O(log 3/4 n).