Graph Bisection Modeled as Binary Quadratic Task Allocation and Solved via Tabu Search

Size: px
Start display at page:

Download "Graph Bisection Modeled as Binary Quadratic Task Allocation and Solved via Tabu Search"

Transcription

1 Graph Bisection Modeled as Binary Quadratic Task Allocation and Solved via Tabu Search Mark Lewis a* and Gary Kochenberger b a Steven Craig School of Business, Missouri Western State University, Saint Joseph, MO b School of Business, University of Colorado, Denver, CO Abstract Balanced graph bisection is an NP-complete problem which partitions a set of nodes in the graph G = (N,E) into two sets with equal cardinality such that a minimal sum of edge weights exists between the nodes in the two separate sets. In this paper we transform graph bisection to capacitated task allocation using variable substitutions to remove most capacity and assignment constraints. The resulting problem is in the form of a generic xqx unconstrained quadratic binary problem, except for a single cardinality constraint. Problems are solved using tabu search employing strategic oscillation around critical events created when the single constraint is satisfied. On a set of benchmark graphs, we improve the best known solution for several problems. Comparison results with the fast, freely available multilevel balanced graph partitioning program METIS are presented on a set of random graphs. Our approach works well when balanced graph bisection is augmented to the 2-processor task allocation problem which adds node preferences for a set, as well as edge weights, to the objective function. For these problems, our approach compares favorably to Cplex and Gurobi, providing better solutions in a much shorter time. Keywords: Combinational analysis; Graph bisection, Heuristics; task allocation * Corresponding author: addresses: mlewis14@missouriwestern.edu, Gary.kochenberger@ucdenver.edu 1 of 17

2 1 Introduction Partitioning a graph G = (N, E) of nodes and edges into k segments is useful in many areas, such as parallel computing implementations, [1], numerical solution of differential equations, [2], designing circuit boards and VLSI design, [3] and data mining, [4]. Balanced graph bisection attempts to partition a graph into two sets each with equal node cardinality, while minimizing total edge costs between the nodes of the two sets. Balanced partitioning of a graph is an NP-complete problem with even small instances difficult to solve exactly. State of the art methods, such as described in [5] are able to exactly solve balanced partitioning problems with up to 90 nodes, while [6] have used semidefinite programming relaxations to find nearly optimal bisections of graphs with 250 nodes. To find good solutions to larger problems, heuristic techniques are employed, for which there are excellent techniques, see for example [7]. Polynomial and logarithmic time bisection algorithms having good approximation ratios are discussed in [8]. Many tabu search and evolutionary algorithms are based on the iterative approach for graph bisection presented in the seminal work of [9]. In practice, the objective of the balanced bisection problem is to create two sets of nodes with a minimum edge cut between them and approximately the same number of nodes in each set. Thus, some programs trade-off having exactly equal node cardinality for achieving a smaller total cut cost, particularly when the number of nodes is very high. 1.1 Multiple Partitioning Multiple partitions of a graph can be created by applying bisection repeatedly to the resulting sets, see [10], [11], [12]. When working with very large graph bisections, multilevel algorithms are employed. Multilevel schemes reduce the original graph to an easily solvable size by combining nodes and edges followed by node/set swap refinements (see [13], [14]). Multilevel algorithms for graph partitioning consist of three phases: coarsening the graph until it is an appropriate size, performing a partition on 2 of 17

3 the coarsened graph, then making adjustments as the coarsened graph is projected backwards to create the desired number of partitions. In a graph that has been coarsened to an appropriate size, a good graph bisection algorithm provides a quick, high quality starting solution for the recombining and refinement portion. Multilevel implementations, such as METIS, see [15], are of near linear complexity. Thus, they are fast and scale well to very large graphs and also have been shown to provide good objective function values. 1.2 Two Processor Task Allocation Hendrickson and Kolda [16] provide a survey of partitioning models showing that the typical objective function for graph partitioning is lacking in expression for the underlying problem and suggest several alternative partitioning models. Alternative models allowing an edge to connect more than two nodes, multi-objective partitioning and skewed partitioning can more accurately model a situation. For example, the mapping of activities in a business process into organizational functional areas having preferences for those activities is very close to a partitioning problem, but more accurately modeled as task allocation. Partitioning social networks where the people (nodes) have preferences for the partition into which they might be placed, or the partitioning of air traffic control sectors into airspace blocks that have preferences for which controllers they utilize are other examples where preferences affect partitioning. Skewed partitioning [16-18] includes in the objective function preference values of nodes for sets, essentially creating a two-processor task allocation problem. Skewed partitioning also has application in modeling VLSI design, optimizing the placement of circuit components and in multi-processor task allocation [18]. For example, in multiprocessor task allocation, a task which is represented by a node, may be math intensive and would best be assigned to a math-friendly processor, but in order to complete the task calculations, it must communicate with a large number of other tasks that are not math intensive, e.g. database or communication intensive tasks. The trade-off between assigning this task to be with the other tasks it communicates with versus assigning it to a 3 of 17

4 compatible processor is more accurately modeled as a task allocation problem rather than a balanced graph bisection [19-21]. The task allocation problem (TAP) minimizes the inter-task communication plus the assignment costs of tasks to processors. TAP is an NP-Hard, binary integer quadratic program [22] appearing in uncapacitated (UTAP) and processor capacitated (CTAP) forms. Balanced graph bisection can be modeled as a CTAP problem involving assigning nodes (tasks) to sets (processors) where the two sets (two processors) have capacity constraints of an equal number of nodes (# tasks divided by two). Common heuristics for TAP include multi-start genetic algorithms [6, 23] and multistart tabu search strategies for binary quadratic optimization [24]. Lewis and Kochenberger [25] use a continuous approximation to a quadratic binary problem to solve CTAP instances up to 100 tasks and 10 processors and they show that as the number of processors is decreased, problem difficulty increases. To our knowledge, we are the first to pursue graph bisection using techniques typically applied to task allocation. As a side note, METIS can solve graph bisections that include node weights. It creates partitions such that the node weights (rather than the count of nodes) are evenly balanced between the two sets. For example, [26] used node weighting to quantify the ability of a cell phone to infect other phones via connections, proposing a balanced partition of a social interaction graph to help contain and treat the infection. However, a single weight on a node does not accurately quantify a preference of a node for inclusion in various sets. In this paper we propose an approach that is applicable to both balanced graph bisection as well as two-processor task allocations (2-TAP). We employ variable substitution to take advantage of the binary nature of these problems, recast them within a generic xqx unconstrained binary quadratic optimization framework, then solve them using a strategic oscillation tabu search algorithm. Using problem instances with up to 3000 nodes as well as a benchmark set of graphs from [27], we compare our approach to METIS [15] and the starting heuristics employed in the commercial solvers Cplex and Gurobi. 4 of 17

5 The remaining sections of this paper are organized as follows. Section 2: presentation of mathematical models. Section 3: discussion of tabu search metaheuristic. Section 4: presentation and discussion of computational experience. Section 5: summary and conclusions. 2 Mathematical Model We first formulate the k-tap model, then 2-TAP, which is an augmentation of balanced graph bisection. Let P = {P 1, P 2,, P m } be the set of m distributed processors (sets in graph partitioning) and T = {T 1, T 2,, T n } be the set of n tasks (nodes) to be run on the processors. Let U = {(i,j) : T i and T j interact at some cost} and let c ij be this edge cost between nodes T i and T j. Let q tp be the assignment cost of task (node) t to processor (set) p. This is the node preference cost that differentiates 2-TAP from graph bisection. Therefore, if all q tp = 0, then the problem is a balanced graph bisection. Let x tp be the binary decision variable that is equal to 1 if node t is assigned to set p, and is otherwise 0. In the initial model, 1 - x tp is denoted by x which indicates that tp task t is not assigned to set p. Each task t is required assignment to one of the processors p. The standard model is: n m Min t 1 p 1 q tp x tp m + c x x ij ip jp i, ju, p 1 i j s.t. m p 1 x tp = 1 for t = 1,, n (1) n t 1 x tp = n / m for p = 1,, m x tp {0, 1} for t = 1,, n and p = 1,, m The first term of the minimization objective function is the preference of a task for a processor and is equal to zero for balanced graph bisection. The second term sums 5 of 17

6 the edge costs between tasks that are not allocated to the same processor, that is, they represent nodes in separate sets with edge costs across the partition. The first constraint requires that each task be assigned to only one processor and the second constraint requires that each processor should have an equal number of tasks for a balanced partition into m processors. Note that partitions may not be perfectly balanced, e.g. if there are an odd number of nodes, then a bisection algorithm adds a dummy node to the graph. By making the variable substitution (1 - x tp ) = x we can sum up the benefits tp associated with assigning two nodes to the same set and subtract that constant from the costs associated with assigning nodes to the same set. In other words, let l tp = q tp + t, t ' U, t t ' c t t be the cost associated with assigning a node t to a set p plus all edge costs between the two sets. Thus the second term in the new objective quantifies the benefits associated with having nodes i and j in the same set p. When the number of partitions is two, i.e., m = 2, we have the 2-TAP problem and the following model is used. This model does not change the number of variables needed but the unconstrained part of the model is a maximum flow problem (see [22]), which should be a benefit for commercial optimization packages with excellent presolvers, such as Cplex. The standard 2-TAP model with m = 2 is: n Min t 1 2 p 1 l tp x tp 2 - c x x ij ip jp i, ju, i j p 1 (2) s.t. x x = 1 for t = 1,, n (3) t1 t 2 n t 1 x tp = n / 2 for p = 1, 2 (4) x tp {0, 1} for t = 1,, n and p = 1, 2 6 of 17

7 Lastly, since a graph bisection considers exactly two sets, the variable substitution x i2 = (1 - x i1 ) dramatically reduces the number of variables needed and eliminates the n assignment constraints (3) and the p capacity constraints (4). A single constraint is now needed to specify that half the binary variables be assigned to the first processor. Except for this single cardinality constraint, the model is now in the form of an xqx unconstrained minimization problem. This reduced model, which we employed in our testing, uses variables with a single subscript to indicate whether a task is assigned to set 1. n Min t 1 l t x t - c x x ij i j i, ju, i j s.t. n t 1 x t = n / 2 x t {0, 1} for t = 1,, n A small example using this reduced model is provided below. Minimize 16 x x x x 4-8 x 1 x 2-4 x 1 x 3-20 x 1 x 4-14 x 2 x 3-28 x 2 x 4-20 x 3 x 4 st x 1 + x 2 + x 3 + x 4 = 2 x t is binary The corresponding Q matrix for this problem is Q = The optimal solution is to partition the graph such that tasks 1 and 4 are in set 1, therefore tasks 2 and 3 are in set 2. The resulting objective value is of 17

8 3 Tabu Search Strategic Oscillation Strategic oscillation constitutes one of the primary strategies of tabu search. The variant of strategic oscillation we employ consists of cycles of varying amplitude around critical events. For the class of problems considered here, a critical event is said to occur when the number of variables equal to one satisfies the cardinality constraint and there is an improvement to the objective function value. The method alternates between constructive phases that progressively set variables to 1 (whose steps we call add moves ) and destructive phases that progressively set variables to 0 (whose steps we call drops moves ). To control the underlying search process, memory structures that are updated at critical events are used. Solutions corresponding to critical events are called critical solutions. A parameter span is used to indicate the amplitude of oscillation about a critical event, where span amplitude is a parameter measured by the number of variables added or dropped in a cycle. We begin with span equal to 1 and gradually increase it to a limiting value. For each value of span, a series of alternating constructive and destructive phases is executed until the span limit is reached. In other words, during a constructive phase all the variables currently equal to 0 are evaluated, one at a time, for their effect on being set to 1 and similarly for the destructive phase. At the upper limit of the current span, variables equal to 1 are selected to be set to 0 during the destructive phase, continuing until the lower span limit is reached. The basic oscillation process is shown in Figure 1. In this set of tests, the number of span cycles around a critical event was set to three and the upper span limit was set to seven. Information stored at critical events is used to influence the tabu search process by generating tabu evaluation criteria based on short term and long term frequency information. For example, a short term "recency" vector is used to record the frequency at which variables have been assigned the value 1 in the most recent set of critical events. Long term memory stores all critical event information and is used to introduce a subtle long term bias into the search process that is useful for breaking ties between candidate variables with equal selection evaluation criteria. A detailed description of the method with pseudocode is given in [28]. In the computational section that follows, we refer to this tabu search heuristic by KL. 8 of 17

9 Figure 1. Oscillation around a critical event # of variables equal to 1 in current solution Start of a destructive phase span cycle n / 2 Critical Event Span limit iterations 4 Computational Results We compare our graph bisection approach with the results of other approaches using a benchmark set of graphs and to METIS, Cplex and Gurobi using a set of randomly generated graphs. Cplex is a widely available, well known and high performance optimization benchmark, which uses branch-and-cut to find exact solutions to mixed integer and quadratic programs. Cplex incorporates pre-solvers to discover useful cliques, networks and variable substitutions prior to starting its search for an exact solution using branch-and-cut. Cplex also employs sophisticated heuristics to quickly construct feasible solutions at node zero of the branch-and-cut tree. Gurboi is a branchand-cut solver that provides free licenses for academic use. Its performance compares favorably with Cplex on MIQP benchmark tests. Gurobi takes advantage of multiple core processors and uses heuristics in its optimization engine. Proving optimality of a graph bisection solution is very difficult, with Cplex and Gurobi able to prove optimality on only the smallest problems, and more often than not, running out of memory before finding solutions of the same quality as our approach. For these problems, Cplex often quickly finds a good solution, then spends a long time in branch-and-cut trying to prove the optimality of that solution, only slightly improving the 9 of 17

10 solution along the way. For example, on a dense thirty node balanced bisection, Cplex heuristics were within 1% of the optimal solution within 0.1 seconds, found the optimal in 2900 seconds and proved optimality in 4500 seconds. For a more reasonable comparison with the other heuristics in this paper, we report the objective values and times of the best heuristic solution found by Cplex and Gurobi. A standard Dell desktop processor having a single core 3.2 GHz processor and 2 GB of RAM using Windows XP was used for all comparisons. Our tabu search parameters used default settings derived from other quadratic binary problems previously solved using our unified xqx model. These problems include set partitioning [29], task allocation [20] and linear ordering [30]. Thus, no problem specific parameter tailoring was used for the results reported in Tables 1, 2 and 4. The primary parameters we employed were span = 7 and cycles = 3, although Table 3 reports solution improvements found when letting the heuristic run for a greater number of cycles and spans. These two parameters also define the stopping criteria for the heuristic. The results reported in the tables from our tabu search method are denoted by KL. 4.1 Bisection Comparisons using Randomly Generated Graphs The problems reported in Table 1 were randomly generated graphs with varying edge densities and no node preference weighting. The objective function is to minimize the total edge weights between two sets having an equal number of nodes. In general, all three solution techniques produced objective function values within a few percent of each other. METIS is a near linear time complexity algorithm and the results verify that it is only minimally impacted by increases in the number of nodes or edge density and is very fast. Table 1 shows that our KL algorithm is faster than Cplex and generally finds a better heuristic solution. KL does not scale to problems with a higher number of nodes as well as METIS. For a given number of nodes, increasing edge density seems to reduce problem difficulty. Table 1a provides a comparison to Gurobi on a representative sample of the problems from Table 1. For these problems, Cplex found a better starting solution faster than Gurobi with KL faster than both to consistently better solutions. The stopping 10 of 17

11 criteria for both Cplex and Gurobi was the incumbent feasible solution at the end of their heuristic search at node zero, in other words, the first node of the start of branch-and-cut. For some of the problems, branch-and-cut was left to run until a normal termination condition was encountered, but it was unable to prove optimality for any of these problems. Table 1 Results from randomly generated graph bisections (no node preferences) # nodes % edge density time (sec) Cplex KL METIS KL time factor improvement over over Cplex METIS KL % solution improvement over over Cplex METIS 50 10% % % % % % % % % % % Table 1a Comparison between KL and Gurobi using selected Table 1 instances. # nodes time (sec) % edge density Gurobi Cplex KL KL time factor improvement over Gurobi over Cplex KL % solution improvement over Gurobi over Cplex % % 0.74% % % 0.34% % % 0.12% % % 1.05% % % 3.0% % % 5.6% 11 of 17

12 4.2 Bisection Results Using Benchmark Graphs In Table 2 we present the results of applying our approach to the benchmark set of graph bisections found in [27]. As mentioned previously, no tailoring of tabu search parameters were applied to obtain these results. The solutions marked with an asterisk are known to be optimal and our approach quickly found these solutions. It should be noted that the best known solutions were often generated using days of computing time, while our times are measured in seconds. Our approach is strongest on those graphs with higher edge densities, weaker on very sparse graphs. For example, we quickly find the best known solution, or a new best solution, on all the densest graphs, but have some trouble with the 500 node and 1000 node graphs with low average node degree. As a side note, Cplex was also used on these problems; however the Cplex results are not reported because the times were not competitive, for example, Cplex ran for 2 hours on the 250 node 8% average node degree problem before running out of memory with an objective function value of 840 (versus 828 in under 4 seconds by KL). Encouraged by our good results on the benchmarks, we did further tests on these graphs to see if we could improve the best known solutions. To this end, we let problems run for extended numbers of cycles, and for larger problems increased the span size. By doing so, we were able to match the best known solution to the 500 node 0.01 degree graph and improve the solutions on the 1000 node graphs with 0.05 and 0.1 average degree (see Table 3). Thus, our approach improved the best known solution for the three largest, densest graphs. 12 of 17

13 Table 2 Results on Benchmark Johnson Graphs # nodes Average node degree Best known solution KL time (sec) Delta from best * * * * * *indicates proven optimal Table 3 Johnson Graph Results with Extended Run Times Average node Previous best KL time Solution # nodes degree known solution solution (min) improvement Comparison Results based on 2-TAP Bisection Model Table 4 compares our results to Cplex on a set of randomly generated 2-TAP graphs of varying edge density, including node preferences for set inclusion. Note that METIS does not solve this category of problem. The table illustrates that our approach generated solutions that improved the Cplex heuristic solution and did so in a shorter period of time. 13 of 17

14 The results also show that our approach scales better as node count and edge density increases. For example we are about 10 times faster for the smaller problems and 100 times faster to a better solution for the largest problems. Table 4a illustrates that for the problems with 2500 nodes and more, the Gurobi optimizer spends less time generating an initial heuristic starting solution than Cplex does, but their starting solution is not as good. KL is about 3x faster than Gurobi on the larger 2-TAP problems and found better solutions to all problems. Based on the typical MIP gap of 100% between incumbent solution and current branch and cut relaxation, for both Gurobi and Cplex, 2-TAP is more difficult than balanced graph bisection. Table 4 Results for the 2-TAP graphs (having edge weights & node preferences) Time (sec) % % solution Cplex Cplex (sec) / KL KL # nodes density improvement* (heuristic) (sec) % 5.0% % 2.8% % 3.0% % 2.1% % 3.0% % 1.6% % 3.8% % 3.2% % 3.2% % 3.0% % 3.5% % 2.6% % 3.1% % 3.0% % 5.6% % 2.6% * percent reduction in objective function of KL over Cplex, computed as 1 KL / Cplex 14 of 17

15 Table 4a Comparison between KL, Cplex and Gurobi using selected Table 4 instances. # nodes time (sec) % edge density Gurobi Cplex KL KL time improvement over Gurobi Over Cplex KL solution improvement Over Gurobi Over Cplex % % 5.0% % % 3.0% % % 1.6% % % 3.1% % % 3.0% % % 5.6% 5 Summary & Conclusions In this paper we augmented the balanced graph bisection model by incorporating preferences between nodes and sets, thus creating capacitated task allocation problems. These 2-TAP problems were modeled as binary quadratic programs. High quality graph bisections were generated using tabu search with strategic oscillation around critical events. Results on a benchmark set of graph bisections were very competitive in solution quality and time to solution and we improved the best known solution for three of the fifteen problems. On a set of randomly generated graph bisection problems, our approach provided better solutions than either METIS, Cplex or Gurobi, although METIS was faster. Our approach to solving 2-TAP was shown to be computationally effective when compared to the heuristically generated solutions of Cplex and Gurobi and our approach scaled better than either as graph size increased. Future research involves integrating our approach as the partitioning phase of a multi-level scheme for k-partitioning much larger graphs. 15 of 17

16 References [1] Ercal F., Ramanujam, J., and Sadayappan, P. Task Allocation on a Hypercube by Recusive Mincut Bipartitioning. Journal of Parallel and Distributed Computing 1990; 10: [2] Hu Y. F. and Blake, R. J. Load Balancing for Unstructured Mesh Applications. Parallel and Distributed Computing Practices 1999; 2: [3] Alpert C. and Kahng, A. Recent directions in netlist partitioning: a survey. INTEGRATION, the VLSI journal 1995; 19: [4] Zha H., He, X., Ding, C., Simon, H., and G, M., "Bipartite graph partitioning and data clustering," presented at Proceedings of the ACM 10th International Conference on Information and Knowledge (CIKM 2001), [5] Billionnet A., Elloumi, S., and Plateau, M.-C. Quadratic Convex Reformulation: a Computational Study of the Graph Bisection Problem. Technical Report CEDRIC, [6] Rendl F., Rinaldi, G., and Wiegele, A. Solving Max-cut to Optimality by Intersecting Semidefinite and Polyhedral Relaxations. Mathematical Programming 2008; 121: [7] Feige U. and Krauthgame, R. A Polylogarithmic approximation of the minimum bisection. SIAM Journal on Computing 2002; 31: [8] Andreev K. and Racke, H., Theory of Computing Systems: Balanced Graph Partitioning, vol. 39, (Eds.): SpringerLink, [9] Kernighan B. W. and Lin., S. An efficient heuristic procedure for partitioning graphs. Bell System Technical Journal 1970; 49: [10] Chen J., Fang, H.-r., and Sadd, Y. Fast Approximate knn Graph Construction for High Dimensional Data via Recursive Lanczos Bisection. Journal of Machine Learning Research 2009; 10: [11] Barnard S. T. and Simon, H. D. Fast multilevel implementation of recursive spectral bisection for partitioning unstructured problems. Concurrency and Computation: Practice and Experience 1994; 6: [12] Simon H. D. and Teng, S. H. How good is recursive bisection? SIAM Journal on Scientific Computing 1995; 18: [13] Chevalier C. and Safro, I., Comparison of Coarsening Schemes for Multilevel Graph Partitioning, vol. 5851, (Eds.): SpringerLink, [14] Karypis G. and Kumar, V. A Fast and High Quality Multilevel Scheme for Partitioning Irregular Graphs. SIAM JOURNAL ON SCIENTIFIC COMPUTING 1999; 20: [15] Karypis G. and Kumar, V., "METIS: A Software Package for Partitioning Unstructured Graphs, Partitioning Meshes, and Computing Fill-Reducing Orderings of Sparse Matrices," V4.0 ed: University of Minnesota, Department of Computer Science / Army HPC Research Center, [16] Hendrickson B. and Kolda, T. G. Graph partitioning models for parallel computing. Parallel Computing 2000; 26: [17] King C. T. Skewed partition-theory and practice [parallel programming]. Proceedings of the Fifteenth Annual International Computer Software and Applications Conference 1991: [18] Hendrickson B., Leland, R., and Driessche, R. V., "Skewed Graph Partitioning," presented at Eighth SIAM Conference Parallel Processing for Scientific Computing, [19] Ernst A., Jiang, H., and Krishnamoorthy, M. Exact Solutions to Task Allocation Problems. Management Science 2006; 52: [20] Lewis M., Alidaee, B., and Kochenberger, G. Using xqx to Model and Solve the 16 of 17

17 Uncapacitated Task Allocation Problem. Operations Research Letters 2005; 33: [21] Menon S. Effective Reformulations for Task Allocation in Distributed Systems with a Large Number of Communicating Tasks. IEEE Transactions on Knowledge and Data Engineering 2004; 16: [22] Billionet A., Costa, M. C., and Sutter, A. An Efficient Algorithm for a Task Allocation Problem. Journal of the Associatlon for Computing Machinery 1992; 39: [23] Hadj-Alounae A. B., Bian, J., and Murty, K. A hybrid genetic/optimization algorithm for a task allocation problem. Journal of Scheduling 1999; 2: [24] Palubeckis G. Multistart Tabu Search Strategies for the Unconstrained Binary Quadratic Optimization Problem. Annals of Operations Research 2004; 131: [25] Lewis M. and Kochenberger, G. A two-stage approach to solving large capacitated task allocation problems. Int. Journal of Mathematical Modelling and Numerical Optimisation 2010; 1: [26] Zhu Z., Cao, G., Zhu, S., Ranjan, S., and Nucci, A. A social network based patching scheme for worm containment in cellular networks. IEEE INFOCOM [27] Johnson D. S., Aragon, C. R., McGeoch, L. A., and Schevon, C. Optimization by simulated annealing: an experimental evaluation. part i, graph partitioning. Operations Research 1989; 37: [28] Glover F., Kochenberger, G., and Alidaee, B. Adaptive memory tabu search for binary quadratic programs. Management Science 1998; 44: [29] Wang H., Kochenberger, G., Lewis, M., Lewis, K., and Alidaee, B. Modeling and Solving Set Packing Problems Via Unconstrained Quadratic Programming. European Journal Of Operational Research 2008; 186: [30] Lewis M., Alidaee, B., Glover, F., and Kochenberger, G. A Note on xqx as a Modeling and Solution Framework for the Linear Ordering problem. International Journal of Operational Research 2009; 5: of 17

A Diversified Multi-Start Algorithm for Unconstrained Binary Quadratic Problems Leveraging the Graphics Processor Unit

A Diversified Multi-Start Algorithm for Unconstrained Binary Quadratic Problems Leveraging the Graphics Processor Unit A Diversified Multi-Start Algorithm for Unconstrained Binary Quadratic Problems Leveraging the Graphics Processor Unit Mark Lewis Missouri Western State University, Saint Joseph, Missouri 64507, USA mlewis14@missouriwestern.edu

More information

Solving the Graph Bisection Problem with Imperialist Competitive Algorithm

Solving the Graph Bisection Problem with Imperialist Competitive Algorithm 2 International Conference on System Engineering and Modeling (ICSEM 2) IPCSIT vol. 34 (2) (2) IACSIT Press, Singapore Solving the Graph Bisection Problem with Imperialist Competitive Algorithm Hodais

More information

Local Search Approximation Algorithms for the Complement of the Min-k-Cut Problems

Local Search Approximation Algorithms for the Complement of the Min-k-Cut Problems Local Search Approximation Algorithms for the Complement of the Min-k-Cut Problems Wenxing Zhu, Chuanyin Guo Center for Discrete Mathematics and Theoretical Computer Science, Fuzhou University, Fuzhou

More information

Multilevel Graph Partitioning

Multilevel Graph Partitioning Multilevel Graph Partitioning George Karypis and Vipin Kumar Adapted from Jmes Demmel s slide (UC-Berkely 2009) and Wasim Mohiuddin (2011) Cover image from: Wang, Wanyi, et al. "Polygonal Clustering Analysis

More information

Native mesh ordering with Scotch 4.0

Native mesh ordering with Scotch 4.0 Native mesh ordering with Scotch 4.0 François Pellegrini INRIA Futurs Project ScAlApplix pelegrin@labri.fr Abstract. Sparse matrix reordering is a key issue for the the efficient factorization of sparse

More information

Parallel Algorithm for Multilevel Graph Partitioning and Sparse Matrix Ordering

Parallel Algorithm for Multilevel Graph Partitioning and Sparse Matrix Ordering Parallel Algorithm for Multilevel Graph Partitioning and Sparse Matrix Ordering George Karypis and Vipin Kumar Brian Shi CSci 8314 03/09/2017 Outline Introduction Graph Partitioning Problem Multilevel

More information

On step fixed-charge hub location problem

On step fixed-charge hub location problem On step fixed-charge hub location problem Marcos Roberto Silva DEOP - Departamento de Engenharia Operacional Patrus Transportes Urgentes Ltda. 07934-000, Guarulhos, SP E-mail: marcos.roberto.silva@uol.com.br

More information

Graph Coloring via Constraint Programming-based Column Generation

Graph Coloring via Constraint Programming-based Column Generation Graph Coloring via Constraint Programming-based Column Generation Stefano Gualandi Federico Malucelli Dipartimento di Elettronica e Informatica, Politecnico di Milano Viale Ponzio 24/A, 20133, Milan, Italy

More information

Scalable Clustering of Signed Networks Using Balance Normalized Cut

Scalable Clustering of Signed Networks Using Balance Normalized Cut Scalable Clustering of Signed Networks Using Balance Normalized Cut Kai-Yang Chiang,, Inderjit S. Dhillon The 21st ACM International Conference on Information and Knowledge Management (CIKM 2012) Oct.

More information

Parallel Multilevel Algorithms for Multi-constraint Graph Partitioning

Parallel Multilevel Algorithms for Multi-constraint Graph Partitioning Parallel Multilevel Algorithms for Multi-constraint Graph Partitioning Kirk Schloegel, George Karypis, and Vipin Kumar Army HPC Research Center Department of Computer Science and Engineering University

More information

Research Article Accounting for Recent Changes of Gain in Dealing with Ties in Iterative Methods for Circuit Partitioning

Research Article Accounting for Recent Changes of Gain in Dealing with Ties in Iterative Methods for Circuit Partitioning Discrete Dynamics in Nature and Society Volume 25, Article ID 625, 8 pages http://dxdoiorg/55/25/625 Research Article Accounting for Recent Changes of Gain in Dealing with Ties in Iterative Methods for

More information

Multi-Objective Hypergraph Partitioning Algorithms for Cut and Maximum Subdomain Degree Minimization

Multi-Objective Hypergraph Partitioning Algorithms for Cut and Maximum Subdomain Degree Minimization IEEE TRANSACTIONS ON COMPUTER AIDED DESIGN, VOL XX, NO. XX, 2005 1 Multi-Objective Hypergraph Partitioning Algorithms for Cut and Maximum Subdomain Degree Minimization Navaratnasothie Selvakkumaran and

More information

My favorite application using eigenvalues: partitioning and community detection in social networks

My favorite application using eigenvalues: partitioning and community detection in social networks My favorite application using eigenvalues: partitioning and community detection in social networks Will Hobbs February 17, 2013 Abstract Social networks are often organized into families, friendship groups,

More information

MULTILEVEL OPTIMIZATION OF GRAPH BISECTION WITH PHEROMONES

MULTILEVEL OPTIMIZATION OF GRAPH BISECTION WITH PHEROMONES MULTILEVEL OPTIMIZATION OF GRAPH BISECTION WITH PHEROMONES Peter Korošec Computer Systems Department Jožef Stefan Institute, Ljubljana, Slovenia peter.korosec@ijs.si Jurij Šilc Computer Systems Department

More information

Graph Partitioning for High-Performance Scientific Simulations. Advanced Topics Spring 2008 Prof. Robert van Engelen

Graph Partitioning for High-Performance Scientific Simulations. Advanced Topics Spring 2008 Prof. Robert van Engelen Graph Partitioning for High-Performance Scientific Simulations Advanced Topics Spring 2008 Prof. Robert van Engelen Overview Challenges for irregular meshes Modeling mesh-based computations as graphs Static

More information

A Computational Study of Conflict Graphs and Aggressive Cut Separation in Integer Programming

A Computational Study of Conflict Graphs and Aggressive Cut Separation in Integer Programming A Computational Study of Conflict Graphs and Aggressive Cut Separation in Integer Programming Samuel Souza Brito and Haroldo Gambini Santos 1 Dep. de Computação, Universidade Federal de Ouro Preto - UFOP

More information

Meshlization of Irregular Grid Resource Topologies by Heuristic Square-Packing Methods

Meshlization of Irregular Grid Resource Topologies by Heuristic Square-Packing Methods Meshlization of Irregular Grid Resource Topologies by Heuristic Square-Packing Methods Uei-Ren Chen 1, Chin-Chi Wu 2, and Woei Lin 3 1 Department of Electronic Engineering, Hsiuping Institute of Technology

More information

Model Reduction for High Dimensional Micro-FE Models

Model Reduction for High Dimensional Micro-FE Models Model Reduction for High Dimensional Micro-FE Models Rudnyi E. B., Korvink J. G. University of Freiburg, IMTEK-Department of Microsystems Engineering, {rudnyi,korvink}@imtek.uni-freiburg.de van Rietbergen

More information

CS 140: Sparse Matrix-Vector Multiplication and Graph Partitioning

CS 140: Sparse Matrix-Vector Multiplication and Graph Partitioning CS 140: Sparse Matrix-Vector Multiplication and Graph Partitioning Parallel sparse matrix-vector product Lay out matrix and vectors by rows y(i) = sum(a(i,j)*x(j)) Only compute terms with A(i,j) 0 P0 P1

More information

Seminar on. A Coarse-Grain Parallel Formulation of Multilevel k-way Graph Partitioning Algorithm

Seminar on. A Coarse-Grain Parallel Formulation of Multilevel k-way Graph Partitioning Algorithm Seminar on A Coarse-Grain Parallel Formulation of Multilevel k-way Graph Partitioning Algorithm Mohammad Iftakher Uddin & Mohammad Mahfuzur Rahman Matrikel Nr: 9003357 Matrikel Nr : 9003358 Masters of

More information

Variable Neighborhood Search for Solving the Balanced Location Problem

Variable Neighborhood Search for Solving the Balanced Location Problem TECHNISCHE UNIVERSITÄT WIEN Institut für Computergraphik und Algorithmen Variable Neighborhood Search for Solving the Balanced Location Problem Jozef Kratica, Markus Leitner, Ivana Ljubić Forschungsbericht

More information

Solving Large Aircraft Landing Problems on Multiple Runways by Applying a Constraint Programming Approach

Solving Large Aircraft Landing Problems on Multiple Runways by Applying a Constraint Programming Approach Solving Large Aircraft Landing Problems on Multiple Runways by Applying a Constraint Programming Approach Amir Salehipour School of Mathematical and Physical Sciences, The University of Newcastle, Australia

More information

Efficient FM Algorithm for VLSI Circuit Partitioning

Efficient FM Algorithm for VLSI Circuit Partitioning Efficient FM Algorithm for VLSI Circuit Partitioning M.RAJESH #1, R.MANIKANDAN #2 #1 School Of Comuting, Sastra University, Thanjavur-613401. #2 Senior Assistant Professer, School Of Comuting, Sastra University,

More information

Randomized rounding of semidefinite programs and primal-dual method for integer linear programming. Reza Moosavi Dr. Saeedeh Parsaeefard Dec.

Randomized rounding of semidefinite programs and primal-dual method for integer linear programming. Reza Moosavi Dr. Saeedeh Parsaeefard Dec. Randomized rounding of semidefinite programs and primal-dual method for integer linear programming Dr. Saeedeh Parsaeefard 1 2 3 4 Semidefinite Programming () 1 Integer Programming integer programming

More information

A Randomized Algorithm for Minimizing User Disturbance Due to Changes in Cellular Technology

A Randomized Algorithm for Minimizing User Disturbance Due to Changes in Cellular Technology A Randomized Algorithm for Minimizing User Disturbance Due to Changes in Cellular Technology Carlos A. S. OLIVEIRA CAO Lab, Dept. of ISE, University of Florida Gainesville, FL 32611, USA David PAOLINI

More information

A Development of Hybrid Cross Entropy-Tabu Search Algorithm for Travelling Repairman Problem

A Development of Hybrid Cross Entropy-Tabu Search Algorithm for Travelling Repairman Problem Proceedings of the 2012 International Conference on Industrial Engineering and Operations Management Istanbul, Turkey, July 3 6, 2012 A Development of Hybrid Cross Entropy-Tabu Search Algorithm for Travelling

More information

Multi-Objective Hypergraph Partitioning Algorithms for Cut and Maximum Subdomain Degree Minimization

Multi-Objective Hypergraph Partitioning Algorithms for Cut and Maximum Subdomain Degree Minimization Multi-Objective Hypergraph Partitioning Algorithms for Cut and Maximum Subdomain Degree Minimization Navaratnasothie Selvakkumaran and George Karypis Department of Computer Science / Army HPC Research

More information

An Integrated Design Algorithm for Detailed Layouts Based on the Contour Distance

An Integrated Design Algorithm for Detailed Layouts Based on the Contour Distance An Integrated Design Algorithm for Detailed Layouts Based on the Contour Distance Jae-Gon Kim and Marc Goetschalckx School of Industrial and Systems Engineering Georgia Institute of Technology Atlanta,

More information

CHAPTER 6 ORTHOGONAL PARTICLE SWARM OPTIMIZATION

CHAPTER 6 ORTHOGONAL PARTICLE SWARM OPTIMIZATION 131 CHAPTER 6 ORTHOGONAL PARTICLE SWARM OPTIMIZATION 6.1 INTRODUCTION The Orthogonal arrays are helpful in guiding the heuristic algorithms to obtain a good solution when applied to NP-hard problems. This

More information

6. Tabu Search 6.1 Basic Concepts. Fall 2010 Instructor: Dr. Masoud Yaghini

6. Tabu Search 6.1 Basic Concepts. Fall 2010 Instructor: Dr. Masoud Yaghini 6. Tabu Search 6.1 Basic Concepts Fall 2010 Instructor: Dr. Masoud Yaghini Outline Tabu Search: Part 1 Introduction Illustrative Problems Search Space Neighborhood Structure Tabus Aspiration Criteria Termination

More information

Research Interests Optimization:

Research Interests Optimization: Mitchell: Research interests 1 Research Interests Optimization: looking for the best solution from among a number of candidates. Prototypical optimization problem: min f(x) subject to g(x) 0 x X IR n Here,

More information

A Modified Inertial Method for Loop-free Decomposition of Acyclic Directed Graphs

A Modified Inertial Method for Loop-free Decomposition of Acyclic Directed Graphs MACRo 2015-5 th International Conference on Recent Achievements in Mechatronics, Automation, Computer Science and Robotics A Modified Inertial Method for Loop-free Decomposition of Acyclic Directed Graphs

More information

Multilevel Algorithms for Multi-Constraint Hypergraph Partitioning

Multilevel Algorithms for Multi-Constraint Hypergraph Partitioning Multilevel Algorithms for Multi-Constraint Hypergraph Partitioning George Karypis University of Minnesota, Department of Computer Science / Army HPC Research Center Minneapolis, MN 55455 Technical Report

More information

A Recursive Coalescing Method for Bisecting Graphs

A Recursive Coalescing Method for Bisecting Graphs A Recursive Coalescing Method for Bisecting Graphs The Harvard community has made this article openly available. Please share how this access benefits you. Your story matters. Citation Accessed Citable

More information

Metaheuristic Optimization with Evolver, Genocop and OptQuest

Metaheuristic Optimization with Evolver, Genocop and OptQuest Metaheuristic Optimization with Evolver, Genocop and OptQuest MANUEL LAGUNA Graduate School of Business Administration University of Colorado, Boulder, CO 80309-0419 Manuel.Laguna@Colorado.EDU Last revision:

More information

Application of Fusion-Fission to the multi-way graph partitioning problem

Application of Fusion-Fission to the multi-way graph partitioning problem Application of Fusion-Fission to the multi-way graph partitioning problem Charles-Edmond Bichot Laboratoire d Optimisation Globale, École Nationale de l Aviation Civile/Direction des Services de la Navigation

More information

Fennel: Streaming Graph Partitioning for Massive Scale Graphs

Fennel: Streaming Graph Partitioning for Massive Scale Graphs Fennel: Streaming Graph Partitioning for Massive Scale Graphs Charalampos E. Tsourakakis 1 Christos Gkantsidis 2 Bozidar Radunovic 2 Milan Vojnovic 2 1 Aalto University, Finland 2 Microsoft Research, Cambridge

More information

Stanford University CS359G: Graph Partitioning and Expanders Handout 1 Luca Trevisan January 4, 2011

Stanford University CS359G: Graph Partitioning and Expanders Handout 1 Luca Trevisan January 4, 2011 Stanford University CS359G: Graph Partitioning and Expanders Handout 1 Luca Trevisan January 4, 2011 Lecture 1 In which we describe what this course is about. 1 Overview This class is about the following

More information

K-Ways Partitioning of Polyhedral Process Networks: a Multi-Level Approach

K-Ways Partitioning of Polyhedral Process Networks: a Multi-Level Approach 2015 IEEE International Parallel and Distributed Processing Symposium Workshops K-Ways Partitioning of Polyhedral Process Networks: a Multi-Level Approach Riccardo Cattaneo, Mahdi Moradmand, Donatella

More information

Multilevel k-way Hypergraph Partitioning

Multilevel k-way Hypergraph Partitioning _ Multilevel k-way Hypergraph Partitioning George Karypis and Vipin Kumar fkarypis, kumarg@cs.umn.edu Department of Computer Science & Engineering, University of Minnesota, Minneapolis, MN 55455 Abstract

More information

VLSI Physical Design: From Graph Partitioning to Timing Closure

VLSI Physical Design: From Graph Partitioning to Timing Closure Chapter Netlist and System Partitioning Original Authors: Andrew B. Kahng, Jens, Igor L. Markov, Jin Hu Chapter Netlist and System Partitioning. Introduction. Terminology. Optimization Goals. Partitioning

More information

Genetic Algorithm for Circuit Partitioning

Genetic Algorithm for Circuit Partitioning Genetic Algorithm for Circuit Partitioning ZOLTAN BARUCH, OCTAVIAN CREŢ, KALMAN PUSZTAI Computer Science Department, Technical University of Cluj-Napoca, 26, Bariţiu St., 3400 Cluj-Napoca, Romania {Zoltan.Baruch,

More information

LOCAL SEARCH FOR THE MINIMUM FUNDAMENTAL CYCLE BASIS PROBLEM

LOCAL SEARCH FOR THE MINIMUM FUNDAMENTAL CYCLE BASIS PROBLEM LOCAL SEARCH FOR THE MINIMUM FUNDAMENTAL CYCLE BASIS PROBLEM Abstract E.Amaldi, L.Liberti, N.Maculan, F.Maffioli DEI, Politecnico di Milano, I-20133 Milano amaldi,liberti,maculan,maffioli @elet.polimi.it

More information

On the Computational Behavior of a Dual Network Exterior Point Simplex Algorithm for the Minimum Cost Network Flow Problem

On the Computational Behavior of a Dual Network Exterior Point Simplex Algorithm for the Minimum Cost Network Flow Problem On the Computational Behavior of a Dual Network Exterior Point Simplex Algorithm for the Minimum Cost Network Flow Problem George Geranis, Konstantinos Paparrizos, Angelo Sifaleras Department of Applied

More information

Q. Wang National Key Laboratory of Antenna and Microwave Technology Xidian University No. 2 South Taiba Road, Xi an, Shaanxi , P. R.

Q. Wang National Key Laboratory of Antenna and Microwave Technology Xidian University No. 2 South Taiba Road, Xi an, Shaanxi , P. R. Progress In Electromagnetics Research Letters, Vol. 9, 29 38, 2009 AN IMPROVED ALGORITHM FOR MATRIX BANDWIDTH AND PROFILE REDUCTION IN FINITE ELEMENT ANALYSIS Q. Wang National Key Laboratory of Antenna

More information

Computational Complexity CSC Professor: Tom Altman. Capacitated Problem

Computational Complexity CSC Professor: Tom Altman. Capacitated Problem Computational Complexity CSC 5802 Professor: Tom Altman Capacitated Problem Agenda: Definition Example Solution Techniques Implementation Capacitated VRP (CPRV) CVRP is a Vehicle Routing Problem (VRP)

More information

Level 3: Level 2: Level 1: Level 0:

Level 3: Level 2: Level 1: Level 0: A Graph Based Method for Generating the Fiedler Vector of Irregular Problems 1 Michael Holzrichter 1 and Suely Oliveira 2 1 Texas A&M University, College Station, TX,77843-3112 2 The University of Iowa,

More information

Partitioning. Course contents: Readings. Kernighang-Lin partitioning heuristic Fiduccia-Mattheyses heuristic. Chapter 7.5.

Partitioning. Course contents: Readings. Kernighang-Lin partitioning heuristic Fiduccia-Mattheyses heuristic. Chapter 7.5. Course contents: Partitioning Kernighang-Lin partitioning heuristic Fiduccia-Mattheyses heuristic Readings Chapter 7.5 Partitioning 1 Basic Definitions Cell: a logic block used to build larger circuits.

More information

Penalized Graph Partitioning for Static and Dynamic Load Balancing

Penalized Graph Partitioning for Static and Dynamic Load Balancing Penalized Graph Partitioning for Static and Dynamic Load Balancing Tim Kiefer, Dirk Habich, Wolfgang Lehner Euro-Par 06, Grenoble, France, 06-08-5 Task Allocation Challenge Application (Workload) = Set

More information

A substructure based parallel dynamic solution of large systems on homogeneous PC clusters

A substructure based parallel dynamic solution of large systems on homogeneous PC clusters CHALLENGE JOURNAL OF STRUCTURAL MECHANICS 1 (4) (2015) 156 160 A substructure based parallel dynamic solution of large systems on homogeneous PC clusters Semih Özmen, Tunç Bahçecioğlu, Özgür Kurç * Department

More information

B553 Lecture 12: Global Optimization

B553 Lecture 12: Global Optimization B553 Lecture 12: Global Optimization Kris Hauser February 20, 2012 Most of the techniques we have examined in prior lectures only deal with local optimization, so that we can only guarantee convergence

More information

Hardware-Software Codesign

Hardware-Software Codesign Hardware-Software Codesign 4. System Partitioning Lothar Thiele 4-1 System Design specification system synthesis estimation SW-compilation intellectual prop. code instruction set HW-synthesis intellectual

More information

Requirements of Load Balancing Algorithm

Requirements of Load Balancing Algorithm LOAD BALANCING Programs and algorithms as graphs Geometric Partitioning Graph Partitioning Recursive Graph Bisection partitioning Recursive Spectral Bisection Multilevel Graph partitioning Hypergraph Partitioning

More information

Multi-Attractor Gene Reordering for Graph Bisection

Multi-Attractor Gene Reordering for Graph Bisection Multi-Attractor Gene Reordering for Graph Bisection Inwook Hwang School of Computer Science & Engineering Seoul National University Sillim-dong, Gwanak-gu, Seoul, 11-744 Korea hwang@soar.snu.ac.kr Yong-Hyuk

More information

Clustering of Microarray data via Clique Partitioning

Clustering of Microarray data via Clique Partitioning Journal of Combinatorial Optimization, 10, 77 92, 2005 c 2005 Springer Science + Business Media, Inc. Manufactured in The Netherlands. Clustering of Microarray data via Clique Partitioning GARY KOCHENBERGER

More information

Shape Optimizing Load Balancing for Parallel Adaptive Numerical Simulations Using MPI

Shape Optimizing Load Balancing for Parallel Adaptive Numerical Simulations Using MPI Parallel Adaptive Institute of Theoretical Informatics Karlsruhe Institute of Technology (KIT) 10th DIMACS Challenge Workshop, Feb 13-14, 2012, Atlanta 1 Load Balancing by Repartitioning Application: Large

More information

Anale. Seria Informatică. Vol. X fasc Annals. Computer Science Series. 10 th Tome 1 st Fasc. 2012

Anale. Seria Informatică. Vol. X fasc Annals. Computer Science Series. 10 th Tome 1 st Fasc. 2012 Anale. Seria Informatică. Vol. X fasc. 1 2012 9 HEURISTIC ALGORITHM FOR GRAPH COLORING BASED ON MAXIMUM INDEPENDENT SET Hilal Al Mara beh, Amjad Suleiman Department Basic Science King Saud bin Abdulaziz

More information

Optimal tour along pubs in the UK

Optimal tour along pubs in the UK 1 From Facebook Optimal tour along 24727 pubs in the UK Road distance (by google maps) see also http://www.math.uwaterloo.ca/tsp/pubs/index.html (part of TSP homepage http://www.math.uwaterloo.ca/tsp/

More information

Optimization of fuzzy multi-company workers assignment problem with penalty using genetic algorithm

Optimization of fuzzy multi-company workers assignment problem with penalty using genetic algorithm Optimization of fuzzy multi-company workers assignment problem with penalty using genetic algorithm N. Shahsavari Pour Department of Industrial Engineering, Science and Research Branch, Islamic Azad University,

More information

Image-Space-Parallel Direct Volume Rendering on a Cluster of PCs

Image-Space-Parallel Direct Volume Rendering on a Cluster of PCs Image-Space-Parallel Direct Volume Rendering on a Cluster of PCs B. Barla Cambazoglu and Cevdet Aykanat Bilkent University, Department of Computer Engineering, 06800, Ankara, Turkey {berkant,aykanat}@cs.bilkent.edu.tr

More information

Parallel Logic Synthesis Optimization for Digital Sequential Circuit

Parallel Logic Synthesis Optimization for Digital Sequential Circuit Kasetsart J. (Nat. Sci.) 36 : 319-326 (2002) Parallel Logic Synthesis Optimization for Digital Sequential Circuit Aswit Pungsema and Pradondet Nilagupta ABSTRACT High-level synthesis tools are very important

More information

Learning a classification of Mixed-Integer Quadratic Programming problems

Learning a classification of Mixed-Integer Quadratic Programming problems Learning a classification of Mixed-Integer Quadratic Programming problems CERMICS 2018 June 29, 2018, Fréjus Pierre Bonami 1, Andrea Lodi 2, Giulia Zarpellon 2 1 CPLEX Optimization, IBM Spain 2 Polytechnique

More information

Unit 5A: Circuit Partitioning

Unit 5A: Circuit Partitioning Course contents: Unit 5A: Circuit Partitioning Kernighang-Lin partitioning heuristic Fiduccia-Mattheyses heuristic Simulated annealing based partitioning algorithm Readings Chapter 7.5 Unit 5A 1 Course

More information

Preclass Warmup. ESE535: Electronic Design Automation. Motivation (1) Today. Bisection Width. Motivation (2)

Preclass Warmup. ESE535: Electronic Design Automation. Motivation (1) Today. Bisection Width. Motivation (2) ESE535: Electronic Design Automation Preclass Warmup What cut size were you able to achieve? Day 4: January 28, 25 Partitioning (Intro, KLFM) 2 Partitioning why important Today Can be used as tool at many

More information

Algorithmic complexity of two defence budget problems

Algorithmic complexity of two defence budget problems 21st International Congress on Modelling and Simulation, Gold Coast, Australia, 29 Nov to 4 Dec 2015 www.mssanz.org.au/modsim2015 Algorithmic complexity of two defence budget problems R. Taylor a a Defence

More information

CAD Algorithms. Circuit Partitioning

CAD Algorithms. Circuit Partitioning CAD Algorithms Partitioning Mohammad Tehranipoor ECE Department 13 October 2008 1 Circuit Partitioning Partitioning: The process of decomposing a circuit/system into smaller subcircuits/subsystems, which

More information

Efficient Programming of Nanowire-based Sublithographic PLAs: A Multilevel Algorithm for Partitioning Graphs

Efficient Programming of Nanowire-based Sublithographic PLAs: A Multilevel Algorithm for Partitioning Graphs Efficient Programming of Nanowire-based Sublithographic PLAs: A Multilevel Algorithm for Partitioning Graphs Vivek Rajkumar (University of Washington CSE) Contact: California Institute

More information

A Binary Integer Linear Programming-Based Approach for Solving the Allocation Problem in Multiprocessor Partitioned Scheduling

A Binary Integer Linear Programming-Based Approach for Solving the Allocation Problem in Multiprocessor Partitioned Scheduling A Binary Integer Linear Programming-Based Approach for Solving the Allocation Problem in Multiprocessor Partitioned Scheduling L. Puente-Maury, P. Mejía-Alvarez, L. E. Leyva-del-Foyo Department of Computer

More information

Research Incubator: Combinatorial Optimization. Dr. Lixin Tao December 9, 2003

Research Incubator: Combinatorial Optimization. Dr. Lixin Tao December 9, 2003 Research Incubator: Combinatorial Optimization Dr. Lixin Tao December 9, 23 Content General Nature of Research on Combinatorial Optimization Problem Identification and Abstraction Problem Properties and

More information

Lecture 19: Graph Partitioning

Lecture 19: Graph Partitioning Lecture 19: Graph Partitioning David Bindel 3 Nov 2011 Logistics Please finish your project 2. Please start your project 3. Graph partitioning Given: Graph G = (V, E) Possibly weights (W V, W E ). Possibly

More information

Metaheuristic Development Methodology. Fall 2009 Instructor: Dr. Masoud Yaghini

Metaheuristic Development Methodology. Fall 2009 Instructor: Dr. Masoud Yaghini Metaheuristic Development Methodology Fall 2009 Instructor: Dr. Masoud Yaghini Phases and Steps Phases and Steps Phase 1: Understanding Problem Step 1: State the Problem Step 2: Review of Existing Solution

More information

A Row-and-Column Generation Method to a Batch Machine Scheduling Problem

A Row-and-Column Generation Method to a Batch Machine Scheduling Problem The Ninth International Symposium on Operations Research and Its Applications (ISORA 10) Chengdu-Jiuzhaigou, China, August 19 23, 2010 Copyright 2010 ORSC & APORC, pp. 301 308 A Row-and-Column Generation

More information

F k G A S S1 3 S 2 S S V 2 V 3 V 1 P 01 P 11 P 10 P 00

F k G A S S1 3 S 2 S S V 2 V 3 V 1 P 01 P 11 P 10 P 00 PRLLEL SPRSE HOLESKY FTORIZTION J URGEN SHULZE University of Paderborn, Department of omputer Science Furstenallee, 332 Paderborn, Germany Sparse matrix factorization plays an important role in many numerical

More information

Hierarchical Multi level Approach to graph clustering

Hierarchical Multi level Approach to graph clustering Hierarchical Multi level Approach to graph clustering by: Neda Shahidi neda@cs.utexas.edu Cesar mantilla, cesar.mantilla@mail.utexas.edu Advisor: Dr. Inderjit Dhillon Introduction Data sets can be presented

More information

α Coverage to Extend Network Lifetime on Wireless Sensor Networks

α Coverage to Extend Network Lifetime on Wireless Sensor Networks Noname manuscript No. (will be inserted by the editor) α Coverage to Extend Network Lifetime on Wireless Sensor Networks Monica Gentili Andrea Raiconi Received: date / Accepted: date Abstract An important

More information

Algorithm XXX: Mongoose, A Graph Coarsening and Partitioning Library

Algorithm XXX: Mongoose, A Graph Coarsening and Partitioning Library 0 Algorithm XXX: Mongoose, A Graph Coarsening and Partitioning Library TIMOTHY A. DAVIS, Texas A&M University WILLIAM W. HAGER, University of Florida SCOTT P. KOLODZIEJ, Texas A&M University S. NURI YERALAN,

More information

Local search heuristic for multiple knapsack problem

Local search heuristic for multiple knapsack problem International Journal of Intelligent Information Systems 2015; 4(2): 35-39 Published online February 14, 2015 (http://www.sciencepublishinggroup.com/j/ijiis) doi: 10.11648/j.ijiis.20150402.11 ISSN: 2328-7675

More information

Machine Learning for Software Engineering

Machine Learning for Software Engineering Machine Learning for Software Engineering Introduction and Motivation Prof. Dr.-Ing. Norbert Siegmund Intelligent Software Systems 1 2 Organizational Stuff Lectures: Tuesday 11:00 12:30 in room SR015 Cover

More information

Solving the Capacitated Single Allocation Hub Location Problem Using Genetic Algorithm

Solving the Capacitated Single Allocation Hub Location Problem Using Genetic Algorithm Solving the Capacitated Single Allocation Hub Location Problem Using Genetic Algorithm Faculty of Mathematics University of Belgrade Studentski trg 16/IV 11 000, Belgrade, Serbia (e-mail: zoricast@matf.bg.ac.yu)

More information

A Hybrid Recursive Multi-Way Number Partitioning Algorithm

A Hybrid Recursive Multi-Way Number Partitioning Algorithm Proceedings of the Twenty-Second International Joint Conference on Artificial Intelligence A Hybrid Recursive Multi-Way Number Partitioning Algorithm Richard E. Korf Computer Science Department University

More information

Grouping Genetic Algorithm with Efficient Data Structures for the University Course Timetabling Problem

Grouping Genetic Algorithm with Efficient Data Structures for the University Course Timetabling Problem Grouping Genetic Algorithm with Efficient Data Structures for the University Course Timetabling Problem Felipe Arenales Santos Alexandre C. B. Delbem Keywords Grouping Genetic Algorithm Timetabling Problem

More information

An Edge-Swap Heuristic for Finding Dense Spanning Trees

An Edge-Swap Heuristic for Finding Dense Spanning Trees Theory and Applications of Graphs Volume 3 Issue 1 Article 1 2016 An Edge-Swap Heuristic for Finding Dense Spanning Trees Mustafa Ozen Bogazici University, mustafa.ozen@boun.edu.tr Hua Wang Georgia Southern

More information

An Efficient Clustering for Crime Analysis

An Efficient Clustering for Crime Analysis An Efficient Clustering for Crime Analysis Malarvizhi S 1, Siddique Ibrahim 2 1 UG Scholar, Department of Computer Science and Engineering, Kumaraguru College Of Technology, Coimbatore, Tamilnadu, India

More information

Place and Route for FPGAs

Place and Route for FPGAs Place and Route for FPGAs 1 FPGA CAD Flow Circuit description (VHDL, schematic,...) Synthesize to logic blocks Place logic blocks in FPGA Physical design Route connections between logic blocks FPGA programming

More information

Data Partitioning. Figure 1-31: Communication Topologies. Regular Partitions

Data Partitioning. Figure 1-31: Communication Topologies. Regular Partitions Data In single-program multiple-data (SPMD) parallel programs, global data is partitioned, with a portion of the data assigned to each processing node. Issues relevant to choosing a partitioning strategy

More information

Study and Implementation of CHAMELEON algorithm for Gene Clustering

Study and Implementation of CHAMELEON algorithm for Gene Clustering [1] Study and Implementation of CHAMELEON algorithm for Gene Clustering 1. Motivation Saurav Sahay The vast amount of gathered genomic data from Microarray and other experiments makes it extremely difficult

More information

for Parallel Matrix-Vector Multiplication? Umit V. C atalyurek and Cevdet Aykanat Computer Engineering Department, Bilkent University

for Parallel Matrix-Vector Multiplication? Umit V. C atalyurek and Cevdet Aykanat Computer Engineering Department, Bilkent University Decomposing Irregularly Sparse Matrices for Parallel Matrix-Vector Multiplication? Umit V. C atalyurek and Cevdet Aykanat Computer Engineering Department, Bilkent University 06533 Bilkent, Ankara, Turkey

More information

Heuristics in MILP. Group 1 D. Assouline, N. Molyneaux, B. Morén. Supervisors: Michel Bierlaire, Andrea Lodi. Zinal 2017 Winter School

Heuristics in MILP. Group 1 D. Assouline, N. Molyneaux, B. Morén. Supervisors: Michel Bierlaire, Andrea Lodi. Zinal 2017 Winter School Heuristics in MILP Group 1 D. Assouline, N. Molyneaux, B. Morén Supervisors: Michel Bierlaire, Andrea Lodi Zinal 2017 Winter School 0 / 23 Primal heuristics Original paper: Fischetti, M. and Lodi, A. (2011).

More information

Exploring Lin Kernighan neighborhoods for the indexing problem

Exploring Lin Kernighan neighborhoods for the indexing problem INDIAN INSTITUTE OF MANAGEMENT AHMEDABAD INDIA Exploring Lin Kernighan neighborhoods for the indexing problem Diptesh Ghosh W.P. No. 2016-02-13 February 2016 The main objective of the Working Paper series

More information

Optimal Sequential Multi-Way Number Partitioning

Optimal Sequential Multi-Way Number Partitioning Optimal Sequential Multi-Way Number Partitioning Richard E. Korf, Ethan L. Schreiber, and Michael D. Moffitt Computer Science Department University of California, Los Angeles Los Angeles, CA 90095 IBM

More information

Heuristic Graph Bisection with Less Restrictive Balance Constraints

Heuristic Graph Bisection with Less Restrictive Balance Constraints Heuristic Graph Bisection with Less Restrictive Balance Constraints Stefan Schamberger Fakultät für Elektrotechnik, Informatik und Mathematik Universität Paderborn Fürstenallee 11, D-33102 Paderborn schaum@uni-paderborn.de

More information

SPATIAL OPTIMIZATION METHODS

SPATIAL OPTIMIZATION METHODS DELMELLE E. (2010). SPATIAL OPTIMIZATION METHODS. IN: B. WHARF (ED). ENCYCLOPEDIA OF HUMAN GEOGRAPHY: 2657-2659. SPATIAL OPTIMIZATION METHODS Spatial optimization is concerned with maximizing or minimizing

More information

Fundamentals of Integer Programming

Fundamentals of Integer Programming Fundamentals of Integer Programming Di Yuan Department of Information Technology, Uppsala University January 2018 Outline Definition of integer programming Formulating some classical problems with integer

More information

An SDP Approach to Multi-level Crossing Minimization

An SDP Approach to Multi-level Crossing Minimization An SDP Approach to Multi-level Crossing Minimization P. Hungerländer joint work with M. Chimani, M. Jünger, P. Mutzel University of Klagenfurt - Department of Mathematics 15th Combinatorial Optimization

More information

Construction of Minimum-Weight Spanners Mikkel Sigurd Martin Zachariasen

Construction of Minimum-Weight Spanners Mikkel Sigurd Martin Zachariasen Construction of Minimum-Weight Spanners Mikkel Sigurd Martin Zachariasen University of Copenhagen Outline Motivation and Background Minimum-Weight Spanner Problem Greedy Spanner Algorithm Exact Algorithm:

More information

Optimizing Architectural Layout Design via Mixed Integer Programming

Optimizing Architectural Layout Design via Mixed Integer Programming Optimizing Architectural Layout Design via Mixed Integer Programming KEATRUANGKAMALA Kamol 1 and SINAPIROMSARAN Krung 2 1 Faculty of Architecture, Rangsit University, Thailand 2 Faculty of Science, Chulalongkorn

More information

Comparing clustering and partitioning strategies

Comparing clustering and partitioning strategies Comparing clustering and partitioning strategies Carlos Afonso, Fábio Ferreira, José Exposto, and Ana I. Pereira Citation: AIP Conference Proceedings 1479, 782 (2012); doi: 10.1063/1.4756254 View online:

More information

Minimizing Clock Domain Crossing in Network on Chip Interconnect

Minimizing Clock Domain Crossing in Network on Chip Interconnect Minimizing Clock Domain Crossing in Network on Chip Interconnect Parag Kulkarni 1, Puneet Gupta 2, Rudy Beraha 3 1 Synopsys 2 UCLA 3 Qualcomm Corp. R&D Outline Motivation The Router Coloring Problem Approaches

More information

Branch-and-Cut and GRASP with Hybrid Local Search for the Multi-Level Capacitated Minimum Spanning Tree Problem

Branch-and-Cut and GRASP with Hybrid Local Search for the Multi-Level Capacitated Minimum Spanning Tree Problem Branch-and-Cut and GRASP with Hybrid Local Search for the Multi-Level Capacitated Minimum Spanning Tree Problem Eduardo Uchoa Túlio A.M. Toffolo Mauricio C. de Souza Alexandre X. Martins + Departamento

More information

VNS-based heuristic with an exponential neighborhood for the server load balancing problem

VNS-based heuristic with an exponential neighborhood for the server load balancing problem Available online at www.sciencedirect.com Electronic Notes in Discrete Mathematics 47 (2015) 53 60 www.elsevier.com/locate/endm VNS-based heuristic with an exponential neighborhood for the server load

More information