Approximating min-sum k -clustering in metric spaces

In many application areas, complex data sets are often represented by some metric space and metric embedding is used to provide a more structured representation of the data. In many of these applications much greater emphasis is put on preserving the local structure of the original space than on maintaining its complete structure. This is also the case in some networking applications where small world phenomena in communication patterns have been observed. Practical study of embedding has indeed involved with finding embeddings with this property. In this paper we initiate the study of local embeddings of metric spaces and provide embeddings with distortion depending solely on the local structure of the space.

We prove that if a metric probability space with a usual concentration property embeds into a finite dimensional Banach space X, then X has a Euclidean subspace of a proportional dimension. In particular this yields a new characterization of weak cotype 2. We also find optimal lower estimates on embeddings of metric spaces with concentration properties into $ l^k_\infty $ , generalizing estimates of Bourgain—Lindenstrauss—Milman, Carl—Pajor and Gluskin.

In this chapter, we discuss the results of two areas intimately related to the main theme of the book. The first one studies (bi-)Lipschitz embeddings of metric spaces into the space forms of nonpositive curvature, Euclidean and hyperbolic spaces, while the main objective of the second area is Lipschitz selections for maps from metric spaces to the space of convex subsets in Rn

The purpose of Chap. 7 is to give a natural definition of generalized solution of non-solvable non-linear equation. In Sect. 7.1 we give the definition of generalized solution of equation with operator which acts in metric spaces. In Sect. 7.2 we give the definition of new concept of near-solution of non-linear operator equation. In Sect. 7.3 and 7.4 the existence, uniqueness, and correctness of generalized solution are proved. In Sect. 7.5 and 7.6 we investigate issues of embedding of metric spaces and extension of operators which are important for the theory of generalized solutions. Sections 7.7 and 7.8 contain examples of operators and describe numerical aspects of computation of generalized solutions. Section 7.9 is devoted to the development of the theory of generalized solutions of equations in uniform spaces and proximity spaces.KeywordsGeneralize SolutionOperator EquationBasic SpaceContinuous OperatorCauchy SequenceThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Numerous sophisticated local algorithm were suggested in the literature for various fundamental problems. Notable examples are the MIS and $$(\Delta +1)$$ -coloring algorithms by Barenboim and Elkin (Distrib Comput 22(5–6):363–379, 2010), by Kuhn (2009), and by Panconesi and Srinivasan (J Algorithms 20(2):356–374, 1996), as well as the $$O\mathopen {}(\Delta ^2)$$ -coloring algorithm by Linial (J Comput 21:193, 1992). Unfortunately, most known local algorithms (including, in particular, the aforementioned algorithms) are non-uniform, that is, local algorithms generally use good estimations of one or more global parameters of the network, e.g., the maximum degree $$\Delta $$ or the number of nodes $$n$$ . This paper provides a method for transforming a non-uniform local algorithm into a uniform one. Furthermore, the resulting algorithm enjoys the same asymptotic running time as the original non-uniform algorithm. Our method applies to a wide family of both deterministic and randomized algorithms. Specifically, it applies to almost all state of the art non-uniform algorithms for MIS and Maximal Matching, as well as to many results concerning the coloring problem (In particular, it applies to all aforementioned algorithms). To obtain our transformations we introduce a new distributed tool called pruning algorithms, which we believe may be of independent interest.

Numerous sophisticated local algorithm were suggested in the literature for various fundamental problems. Notable examples are the MIS and (Δ+1)-coloring algorithms by Barenboim and Elkin [6], by Kuhn [22], and by Panconesi and Srinivasan [33], as well as the OΔ2-coloring algorithm by Linial [27]. Unfortunately, most known local algorithms (including, in particular, the aforementioned algorithms) are non-uniform, that is, they assume that all nodes know good estimations of one or more global parameters of the network, e.g., the maximum degree Δ or the number of nodes n.This paper provides a rather general method for transforming a non-uniform local algorithm into a uniform one. Furthermore, the resulting algorithm enjoys the same asymptotic running time as the original non-uniform algorithm. Our method applies to a wide family of both deterministic and randomized algorithms. Specifically, it applies to almost all of the state of the art non-uniform algorithms regarding MIS and Maximal Matching, as well as to many results concerning the coloring problem. (In particular, it applies to all aforementioned algorithms.)To obtain our transformations we introduce a new distributed tool called pruning algorithms, which we believe may be of independent interest.

We initiate the study of the computational complexity of the covering radius problem for point lattices, and approximation versions of the problem for both lattices and linear codes. We also investigate the computational complexity of the shortest linearly independent vectors problem, and its relation to the covering radius problem for lattices. For the covering radius on n-dimensional lattices, we show that the problem can be approximated within any constant factor /spl gamma/(n) > 1 in random exponential time 2/sup O(n)/, it is in AM for /spl gamma/(n) = 2, in coAM for /spl gamma/(n) = /spl radic/(n log n), and in NP /spl cap/ coNP for /spl gamma/(n) = /spl radic/n. For the covering radius on n-dimensional linear codes, we show that the problem can be solved in deterministic polynomial time for approximation factor /spl gamma/(n) = log n, but cannot be solved in polynomial time for some /spl gamma/(n) = /spl Omega/(log log n) unless NP can be simulated in deterministic n/sup O(log log log n)/ time. Moreover, we prove that the problem is NP-hard for every constant approximation factor, it is /spl Pi//sub 2/-hard for some constant approximation factor, and it is in AM for approximation factor 2. So, it is unlikely to be /spl Pi//sub 2/-hard for approximation factors larger than 2. This is a natural hardness of approximation result in the polynomial hierarchy. For the shortest independent vectors problem, we give a coAM protocol achieving approximation factor /spl gamma/(n) = /spl radic/(n/log n), solving an open problem of Blomer and Seifert (1999), and prove that the problem is also in coNP for /spl gamma/(n) = /spl radic/n. Both results are obtained by giving a gap-preserving nondeterministic polynomial time reduction to the closest vector problem.

Minimizing the total length has been one of the main optimality criteria for triangulations and other kinds of partitions. Indeed, the minimum weight triangulation or MWT (i.e. a triangulation of minimum total edge length) has frequently been referred to as the optimal This triangulation has some good properties [1] and is e.g. useful in numerical approximation of bivariate data [23]. Its complexity is one of the most intriguing problems in computational geometry. No proof has yet been presented showing that it is NP-hard. On the other hand, no polynomial time heuristic for it was shown to guarantee even a constant approximation factor. However, very recently we proved that what we called the quasi-greedy triangulation has length within a constant factor from the minimum [13]. A standard implementation of that heuristic using known results would yield a quadratic time algorithm. In this paper, we improve on this by describing an O(n log n) algorithm for quasi-greedy triangulation. Moreover, if the Delaunay triangulation is given, our algorithm runs in linear time. The minimum weight convex partition (MWCP) problem has been shown to be NP-complete [7, 17]. By a MWCP we mean a set of non-crossing diagonals (i.e. without Steiner points) of minimum total length which includes the convex hull and partitions it into empty convex polygons. In [13] it was shown that, given the quasi-greedy triangulation, a convex partition of length within a constant factor from the MWCP can be computed in linear time. Thus from the results in this paper we also obtain a fast heuristic for approximating the MWCP. The quasi-greedy triangulation is very closely related to the greedy triangulation (both are defined in the next section). In fact, we will describe an algorithm that, given the Delannay triangulation, computes the greedy triangulation in linear time. This algorithm behaves in such a way that it is very easily adapted to yield an algorithm for the quasi-greedy triangulation. Consequently, we also present the first O(n log n) algorithm for greedy triangulation. (By using a substantially different approach, an attempt to develop an O(n log n) algorithm for greedy triangulation was done in [21]. Even if that approach could be corrected

Lokshtanov, Marx, and Saurabh SODA 2011 proved that there is no (k-epsilon)^pw(G)poly(n) time algorithm for deciding if an n-vertex graph G with pathwidth pw admits a proper vertex coloring with k colors unless the Strong Exponential Time Hypothesis (SETH) is false, for any constant epsilon>0. We show here that nevertheless, when k>lfloor Delta/2 rfloor + 1, where Delta is the maximum degree in the graph G, there is a better algorithm, at least when there are few colorings. We present a Monte Carlo algorithm that given a graph G along with a path decomposition of G with pathwidth pw(G) runs in (lfloor Delta/2 rfloor + 1)^pw(G)poly(n)s time, that distinguishes between k-colorable graphs having at most s proper k-colorings and non-k-colorable graphs. We also show how to obtain a k-coloring in the same asymptotic running time. Our algorithm avoids violating SETH for one since high degree vertices still cost too much and the mentioned hardness construction uses a lot of them. We exploit a new variation of the famous Alon--Tarsi theorem that has an algorithmic advantage over the original form. The original theorem shows a graph has an orientation with outdegree less than k at every vertex, with a different number of odd and even Eulerian subgraphs only if the graph is k-colorable, but there is no known way of efficiently finding such an orientation. Our new form shows that if we instead count another difference of even and odd subgraphs meeting modular degree constraints at every vertex picked uniformly at random, we have a fair chance of getting a non-zero value if the graph has few k-colorings. Yet every non-k-colorable graph gives a zero difference, so a random set of constraints stands a good chance of being useful for separating the two cases.

<p>In combinatorial optimization, covering problems are those problems where given a ground set and a family of subsets of the ground set, the objective is to find a minimum cost set of subsets whose union contains the ground set. We consider covering problems in the context of Computational Geometry, which is a subfield of Computer Science that deals with problems associated with geometric objects such as points, lines, disks, polygons etc. In particular, geometric covering is an active research area, where the ground set and the family of subsets are induced by geometric objects. Covering problems in combinatorial optimizations often have a geometric analogue that arises naturally, and though such problems remain NP-hard, it is often possible to exploit the geometric properties of the set system to obtain better approximation algorithms.</p> <p>In this work, the fundamental problem that we consider is a generalization of geometric covering, where each element in the ground set may need to covered by more than one subset. To be precise, the problem is defined as follows: given two sets of points X, Y and a coverage function κ : X → Z+ ∪ {0}, construct balls centered on the points in Y such that each point in X is covered by at least κ(x) distinct balls. The objective in this problem is to minimize the total cost, which is a function of the radii of the balls. This problem is termed as the metric multi-cover (MMC) problem.</p> <p>We first consider version of the MMC problem where κ(x) = 1 for all clients, i.e. the 1-covering case. The known results that give a constant factor approximation for this problem are variations of LP-based primal-dual algorithm. We use a modified local search technique, motivated by geometric idea, to derive a simple, constant- factor approximation for this problem in Chapter 2.</p> <p>We then look at the MMC problem where the point sets X,Y are in the Euclidean plane, and each client x ∈ X needs to be covered by at least κ(x) distinct disks centered on the points in Y . In Chapter 4, we give the first polynomial time constant factor approximation for this problem, in which the constant is independent of the coverage function κ. Our solution also has an incremental property, which allows the algorithm to handle increases in the coverage requirement by increasing the radii of the current server disks, without affecting the approximation factor.</p> <p>In the next problem, we move from the Euclidean plane to arbitrary metric spaces where we consider the uniform MMC problem. In this problem, each client x has the demand κ(x) = k, where k > 0 is an integer. We give the first constant factor approximation (independent of k) for this problem. The key contribution that led to this result is the formulation of a partitioning scheme of the servers in the uniform MMC problem, that reduces the uniform MMC problem to k instances of 1-covering problem, while preserving the optimality of the solution to a constant multiplicative factor. We present the partitioning scheme as an independent result in Chapter 5, which we then use to solve the uniform MMC problem in Chapter 6.</p> <p>The MMC problem with arbitrary coverage function κ is then considered in Chapter 7. The key challenge that the non-uniform version presents is that the symmetry of the server partitioning scheme breaks down as the coverage demands of clients are independent of each other. We present a constant factor algorithm for this problem in Chapter 7.</p> <p>The last problem that we consider is the t-MMC problem, which is a restricted version of the uniform MMC problem. The objective is to compute a cover in which each client is covered by at least k distinct server disks, using atmost t server disks in total. This problem is a generalization of the clustering problem (where k = 1), and to our knowledge this is the first time this generalization has been considered. We give a constant factor approximation for this problem in Chapter 8.</p>

Semantic and instance segmentation algorithms are two general yet distinct image segmentation solutions powered by Convolution Neural Network. While semantic segmentation benefits extensively from the end-to-end training strategy, instance segmentation is frequently framed as a multi-stage task, supported by learning-based discrimination and post-process clustering. Independent optimizations on substages instigate the accumulation of segmentation errors. In this work, we propose to embed prior clustering information into an embedding learning framework FCRNet, stimulating the one-stage instance segmentation. FCRNet relieves the complexity of post process by incorporating the number of clustering groups into the embedding space. The superior performance of FCRNet is verified and compared with other methods on the nucleus dataset BBBC006.

Answering an open question published in Operations Research (54, 73–91, 2006) in the area of network design and logistic optimization, we present the first constant-factor approximation algorithms for the problem combining facility location and cable installation in which capacity constraints are imposed on both facilities and cables. We study the problem of designing a minimum cost network to serve client demands by opening facilities for service provision and installing cables for service shipment. Both facilities and cables have capacity constraints and incur buy-at-bulk costs. This Max SNP-hard problem arises in diverse applications and is shown in this paper to admit a combinatorial 19.84-approximation algorithm of cubic running time. This is achieved by an integration of primal-dual schema, Lagrangian relaxation, demand clustering and bi-factor approximation. Our techniques extend to several variants of this problem, which include those with unsplitable demands or requiring network connectivity, and provide constant-factor approximate algorithms in strongly polynomial time.

Fast computation of two-dimensional layouts of hierarchically clustered networks is a well-studied problem in graph visualization. We present algorithmic and experimental advances on the subject: (i) We propose a new drawing algorithm that combines space-filling and fast force-directed methods; it runs in O(nlogn+m) time, where n and m are the number of vertices and edges of the network, respectively. This running time does not depend on the number of clusters, thus the algorithm guarantees good time performances independently of the structure of the cluster hierarchy. As a further advantage, the algorithm can be easily parallelized. (ii) We present an experimental analysis aimed at understanding which clustering algorithms can be used, in combination with our visualization technique, to generate better quality drawings for medium and large networks with small-world and scale-free structure. As far as we know, no previous similar experiments have been done in this respect.

Military communication involves sharing of secret information which needs to be done with more concern to assure secured transmission of data. In the war field communication would be more difficult task where the soldiers are scattered in different location. In this case route path establishment and routing would be more difficult task due to presence of scattered soldiers in multiple locations. There is various research works has been focused to provide the efficient and better route path establishment in MANET. In the previous work, load and energy aware micro–macro density clustering approach is introduced which focus on performing clustering in two levels micro clustering and macro clustering. However this research work might degrades its clustering performance with respect to energy and time in case of more cluster count generation and also optimal route path construction is also not performed correctly. This is resolved in proposed research work by introducing new framework called multi-objective aware micro–macro clustering using hybridized additive weight based genetic algorithm (MO–MMC–HAWGA). This proposed research work leads to optimal clustering of mobile node under multiple objectives consideration like distance, energy, bandwidth, and stability. Initially it first chooses the optimal cluster head based on which optimal clustering would be done. This work improves the performance by avoiding the more number of clusters which are generated at run time by performing macro clustering. The research work/s experimental evaluation is conducted in NS2 simulation environment and it is proved that proposed research work namely MO–MMC–HAWGA provides optimal result than existing research methods.

K-means algorithm is one of the most widely used methods in data mining and statistical data analysis to partition several objects in K distinct groups, called clusters, on the basis of their similarities. The main problel and distributed clustering algorithms start to be designem of this algorithm is that it requires the number of clusters as an input data, but in the real life it is very difficult to fix in advance such value. In this work we propose a parallel modified K-means algorithm where the number of clusters is increased at run time in a iterative procedure until a given cluster quality metric is satisfied. To improve the performance of the procedure, at each iteration two new clusters are created, splitting only the cluster with the worst value of the quality metric. Furthermore, experiments in a multi-core CPUs based environment are presented.