L. Sanchis, "Multiple-way Network Partitioning," IEEE Trans. on Computers, pp. 62-81, Vol. 38, Issue 1, 1989.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....functions that depend on knowing how the hyperedges are partitioned across all k partitions has been recognized by various researchers, and a number of di erent algorithms have been developed to minimize objective functions such as the sumof external degrees, scaled cost, absorption etc. [21, 5, 7, 17, 25]) Moreover, direct k way partitioning can potentially produce much better solutions than a method that computes a k way partitioning via recursive bisection. In fact, in the context of a certain classes of graphs it was shown that recursive bisectioning can be up to an O(log n) factor worse than ....

.... objectives, and the theoretical results which suggest that it can lead to superior partitioning solutions, a number of studies have shown that existing direct k way partitioning algorithms for hypergraphs, produce solutions that are in general inferior to those produced via recursive bisectioning [21, 7, 17, 25]. The primary reason for that is the fact that computationally ecient k way partitioning re nement algorithms are often trapped into local minima, and usually require much more sophisticated and expensive optimizers to climb out of them. To overcome these con icting requirements and ....

L. Sanchis. Multiple-way network partitioning. IEEE Trans. On Computers, 38(1):62-81, 1989.

....performed at level i, hypergraph H i is modi ed. We then run iterative local search methods (line 1a) on this new hypergraph H i to calculate new hyperedgecuts as well as to discover potential elite solutions (line 1b) As local search methods we have used the Sanchis partitioning algorithm (FMS) [15], and the multiway partitioning by free moves (PFM) proposed by Dasdan and Aykanat [5] Next, on line 2, the interpolation operator is applied to get an elite solution x from X i 1 , so that x can be used as an initial solution for the search methods FMS and PFM. Finally, the local clustering ....

L.A. Sanchis. Multiple-way Network Partitioning. IEEE Trans. Comput., 38(1):62-81, Jan. 1989.

....Algorithms. Refining a k way partitioning is significantly more complicated because vertices can move from a partition to many other partitions; thus, increasing the optimization space combinatorially. An extension of the FM refinement algorithm in the case of k way refinement is described in [Sanchis, 1989]. This algorithm uses k(k 1) priority queues, one for each type of move. In each step of the algorithm, the moves with the highest gain are found from each of these k(k 1) queues, and the move with the highest gain that preserves or improves the balance, is performed. After the move, all of ....

Sanchis, L. A. (1989). Multiple-way network partitioning. IEEE Transactions on Computers, pages 62--81.

....technique was introduced by Kernighan and Lin [Ker70] By using pair swap moves, the algorithm is able to iteratively move elements from partition to partition. Many different improvements on the strategy have been proposed. The most well known being by Fiduccia and Mattheyses [Fid82] Sanchis [San89], and Krishnamurthy [Kri84] Alpert and Kahng give a survey on the topic [Alp95] Until recently many of the probabilistic algorithms proposed were probabilistic iterative improvement algorithms. Two of the main techniques are the Metropolis algorithm [Met53] and Simulated Annealing [Kir83, ....

L.A. Sanchis. Multiple-way network partitioning. IEEE Transactions on Computers, Vol.38, No. 1, pp.62-81, 1989.

....the degree of cluster C h . Notice that f counts every cut edge twice; we do this to avoid 1 2 terms throughout this work. Min cut graph partitioning is known to be NP complete, so heuristic methods must be invoked. Previous approaches have included seeded epitaxial growth, iterative improvement [16], genetic algorithms [6] etc. Spectral methods [1] 2] 4] 7] 8] 11] 13] 15] have been successful in recent years and are of particular interest for our present work. These works share a common trait of using eigenvectors to construct some type of geometric representation of the graph. We ....

L. A. Sanchis, "Multiple-way Network Partitioning", IEEE Trans. on Computers, 38, 1989, pp. 62-81.

....for k way partitioning by repetitive use of bipartitioning. Since these methods neglect the special properties inherent in problems involving multiple partitions, their results are not as good as the heuristics that operate on multiple partitions simultaneously. The heuristic presented in [SanL89] handles multiple partitions forms the basis of the partitioning algorithm used by the proposed CAE tools. The nature of FPGA based computing arrays provides some additional prospects for optimizing spatial partitions. Unlike VLSI applications, where each additional piece of circuitry has a ....

L. A. Sanchis, "Multiple-Way Network Partitioning," IEEE Transactions on Computers, vol. 38, no. 1, Jan. 1989.

....The iterative improvement algorithms perform a sequence of bi partitioning. During each bi partitioning, either a vertex exchange scheme or a one way vertex moving scheme is used to iteratively reduce the interprocessor communication subject to some constraints on processor workload [28, 52] In [87], Sanchis applied iterative improvements to multiple way partitioning with a more frequent information updating after each vertex exchange or moving and obtained better results. Nandy and Loucks applied iterative improvements to the initial random partitioning for parallel logic simulation [76] ....

....That is, candidate subgraphs are selected based on the gain in edge cut by moving them or exchanging them. In addition, the moving or exchanging of subgraphs must satisfy a lower bound ( 95 of averaged load) and an upper bound ( 105 of averaged load) of processor workload. Previous algorithms [52, 28, 87] that are based on move based methods only allow one way move or two node exchange scheme. They do not allow many to many exchanges that are used in the iCPP algorithm. As a result, when two subgraphs 33 cannot be exchanged without violating workload constraints, the subgraph with larger workload ....

Laura A. Sanchis, "Multiple-Way Network Partitioning," IEEE Transactions on Computers, Vol. 38, No. 1, pp. 62-81, January 1989.

....elements to different partitions. It proceeds in a series of passes, during which each component is moved exactly once. A number of improvements on the basic strategies have been proposed over the years [Gav72] The most notable are the ones proposed by Fiduccia and Mattheyses [Fid82] Sanchis [San89], and Krishnamurthy [Kri84] An excellent survey of this research is given in [Alp95] The iterative improvement paradigm has been applied to many other optimization problems, including the Traveling Salesman Problem [Lin73] Randomized versions of the deterministic constructive algorithms have ....

L.A. Sanchis. "Multiple-way network partitioning." IEEE Transactions on Computers, vol.38(1), pp.62-81, 1989.

....any nodes between blocks manually) their task was to select and rank the most promising pairs of blocks for subsequent focused searches. For our experiments we obtained 8 way partitions of the ibm04, ibm08 networks from the benchmark set in [1] by running the Sanchis refinement algorithm ([5]) on the best solution produced by 200 runs of the hMeTis system [4] one of the most widely used network partitioning systems. We computed a score 1 A more useful search might focus on these nodes and their close neighbors in the hypergraph structure. Our interface allows the user to expand the ....

L. A. Sanchis. Multiple-way network partitioning. IEEE Trans. on Comp., 38:62--81, 1989.

....sizes and nonuniform vertex weights. Partition balance is enforced by bounds on the partition sizes. Variants to the KL and FM algorithms continue to appear in the graph and hypergraph partitioning literature. Krishnamurthy (1984) introduced the concept of level gains into the FM heuristics. Sanchis (1989) adapted Krishnamurthys bipartition algorithm to solve the hypergraph K partitioning problem. Hendrickson and Leland (1993a) have implemented an iterative improvement graph partitioning algorithm patterned after the FM algorithm but generalized in several ways: First, K partitioning is possible. ....

Sanchis, L.A., 1989, Multiple-Way Network Partitioning, IEEE Transactions on Computers, Vol. 38, No. 1, pp. 62---81.

....the relative advantage of MAs over EAs is quite consistent on complex search spaces. Figure 4 shows one possible implementation of a Memetic algorithm based on the generic Genetic Algorithm introduced earlier in Section 3.3. We use a simple variation of the Fiduccia and Mattheyses (FM) heuristic [17]. The original FM heuristic has several passes after which the heuristic terminates as presented in Section 2. In the local optimization phase, a single pass is allowed,furthermore a restriction on the number of modules to be moved is set to a certain value. It is to be noted that if local ....

L.A. Sanchis, "Multiple-Way Network Partitioning ", IEEE Transactions on Computers, vol. 38, n. 1, pp. 62--81, January 1989.

....Lim Research and Teaching Statement 2 ffl KPM (K way Pairwise Movement) is a cutsize driven multiway partitioning algorithm. KPM is an extension of LR bipartitioning into multiway partitioning. We make the first important observation that the well known cutsize driven multiway partitioner KFM [13, 14] is not suitable for today s large scale circuits. Then, we propose a simple yet effective hill climbing method PM (Pairwise Movement) that overcomes the limitation of KFM and provides partitioners a capability to explore wider range of solution space effectively. ffl FLARE is a delay driven ....

L. A. Sanchis. Multiple-way network partitioning. IEEE Trans. on Computer, pages 62--81, 1989.

....algorithms runs in time O(L P) per pass) where L is the number of levels used. Suaris and Kedem [59] extended the FM algorithm to hypergraph quadrisection. Barnes et al. 5] determined sets of vertices to interchange between partitions from solving a transportation (linear) problem. Sanchis [51, 52] adapted Krishnamurthy s bipartition algorithm to the hypergraph K partitioning problem. The time complexity of Sanchis algorithm is O(L P K (log K L D max ) per pass) where D max is the maximum vertex degree in the hypergraph. Hendrickson and Leland [32] have implemented FM HL, an ....

L.A. Sanchis, "Multiple-Way Network Partitioning," IEEE Transactions on Computers, Vol. 38, No. 1, pp. 62--81, 1989.

.... computes a bipartitioning of the nets, and then transforms the net partitioning solution into a module partitioning solution [20, 9] The multi way partitioning algorithms include the recursive bipartitioning by Kernighan and Lin [22] a generalization of the FM algorithm with lookahead by Sanchis [27], the primal dual algorithm [30] and a generalization of the graph spectral based partitioning method to multi way ratio cut by Chan, Schlag, and Zien [5] To reduce the computational complexity for partitioning very large circuits, cluster based partitioning methods have been introduced. In this ....

....K DualFM, consists of the following phases: 1) We first convert the net hypergraph to a dual net representation. The dual net representation used in K DualFM, named hybrid dual netlist (HDN ) is a combination of NIG and DNHG . 2) Assign nets to partitions. We use the K FM partitioning algorithm [27, 15] to partition the nets into K partitions. 3) We transform the net partitioning solution into a module partitioning solution by solving the K way module contention problem (K MC) We show that the K MC problem can be formulated as a min cost max flow problem and we present two efficient algorithms ....

L. Sanchis, "Multiple-Way Network Partitioning," IEEE Trans. on Computers, Vol. 38, pp. 62-81, 1989.

....two goals are considered to reduce cost. First, the number of chips to be used is as small as possible. Second, the total length of interconnections across chip boundaries (i.e. the external interconnection cost) is minimized. This problem is similar to the multiple way graph partitioning problem [KL70, San89, YC91, Kri84, BHJL89]. The difference is that firstly we need to consider one additional goal minimizing the number of partitions, and secondly, a partition should correspond to a chip from the chip library. Our approach is described in Chapter 2. We model the problem as a graph partitioning problem. We first use ....

L. A. Sanchis, "Multiple-way network partitioning", IEEE Trans. Comput., vol. 38, no. 1, pp. 62-81, Jan. 1989.

....instead of vertex swap. This modification as well as proper data structures, e.g. bucket lists, reduced the time complexity of a single pass of KL algorithm to linear in the size of the graph. Here, size refers to the number of edges and pins in a graph and hypergraph, respectively. Sanchis [11] proposed a multiway hypergraph partitioning algorithm which directly handles the partitioning of a hypergraph into more than two parts. Note that all the previous approaches before Sanchis algorithm (SN algorithm) are originally bipartitioning algorithms. As mentioned earlier, all KL based ....

Sanchis, L. A., "Multiple-way network partitioning, " IEEE Trans. on Computers, vol. 38(1), pp. 62--81, 1989.

No context found.

L. Sanchis, "Multiple-way Network Partitioning," IEEE Trans. on Computers, pp. 62-81, Vol. 38, Issue 1, 1989.

No context found.

L. A. Sanchis, Multiple-way network partitioning. In IEEE Trans. Computer, vol. 38, pp.62-81, Jan. 1989.

No context found.

L.A. Sanchis, Multiple-way network partitioning, IEEE Trans. on Computers 38 (1) (1989) 62--81.

No context found.

Laura A. Sanchis. Multiple-Way Network Partitioning. IEEE Transactions on Computers, Volume 38, Number 1, Pages 62-74, January 1989.

No context found.

Sanchis, L., "Multiple-Way Network Partitioning," in IEEE Transactions on Computers, 38, 1989, pp. 62-81.

No context found.

L. A. Sanchis, Multiple-way network partitioning, IEEE Transactions on Computers, 38 (1989), pp. 62--81.

No context found.

L. Sanchis, "Multiple-way Network Partitioning," IEEE Trans. on Computers, pp. 62-81, Vol. 38, Issue 1, 1989.

No context found.

L.A. Sanchis. Multiple-way network partitioning. IEEE Trans. on Comp., 38(1):62--81, 1989.

No context found.

Laura A. Sanchis, Multiple-Way Network Partitioning, IEEE Trans. on Computers, VOL. 38, No.1, Jan. 1989.

First 50 documents  Next 50

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC