Multiple-way network partitioning

Abstract

We present an algorithm for partitioning the cells of a network into an arbitrary number of segments based on a recent 2-way network partitioning algorithm by B. Krishnamurthy [4]. By efficient use of data structures the complexity of the algorithm is shown to increase only linearly in the number of segments in the majority of cases. Through theoretical and experimental methods we show that the concept of "level gain" introduced in [4] becomes more useful as the number of segments increases.

Hybrid Metaheuristics for the Graph Partitioning Problem

“…Several adaptations of the KL and FM procedures have been proposed in the literature for the k-way partitioning. Among these is an extended FM algorithm by Sanchis [38,39], and an adaptation of FM proposed by Hendrickson and Leland [23]. A description of the KL procedure and its modification by Fiduccia and Mattheyses is provided in Section 4.1.…”

“…Several adaptations of the FM algorithm have been proposed for the k-way partitioning. In [38], Sanchis proposes maintaining k(k − 1) previously described bucket structures, one for each of the k(k − 1) possible directions to move a cell Fig. 1 The bucket sorting data structure [21] for graph bisection (i.e., vertex) between partition subsets.…”

Genetic Algorithm Based Approach To Circuit Partitioning

“…A new data structure bucket list for cell gains and proposed cell move with better time complexity was proposed [4]. Krishnamurthy [5] modified [4] to introduce the concept of look ahead to choose the cell move.…”

“…Various multiway partitioning algorithms were proposed by modifying [4] [5] and developing appropriate data structures [6], top down clustering and iterative primal-dual approach [7], dual intersection graph representation and ratio cut metric [8]. Areibi and Vannelli [9] described the Manuscript received August 14, 2009.…”

An Efficient Semidefinite Programming Relaxation for the Graph Partition Problem

“…The graph partition problem has many applications such as VLSI design [36], parallel computing [6,26,44], network partitioning [19,43], and floor planing [9]. The graph equipartition problem also plays a role in telecommunications, see e.g., [37].…”