High-Quality Hypergraph Partitioning
نویسندگان
چکیده
Hypergraphs are a generalization of graphs where edges (aka nets ) allowed to connect more than two vertices. They have similarly wide range applications as graphs. This article considers the fundamental and intensively studied problem balanced hypergraph partitioning (BHP) , which asks for vertices into k disjoint blocks bounded size while minimizing an objective function over hyperedges. Here, we consider most commonly used objectives: cut-net metric connectivity . We describe our open-source partitioner KaHyPar is based on successful multi-level approach—driving it extreme using one level (almost) every vertex. Using carefully designed data structures dynamic update techniques, this approach turns out very good time–quality tradeoff. present preprocessing techniques— pin sparsification locality-sensitive hashing (LSH) community detection Louvain algorithm The structure guide coarsening process that incrementally contracts Portfolio-based contracted then already achieves initial solution. While reversing contraction process, combination several refinement techniques final partitioning. In particular, support highly-localized local search can directly produce -way complement with flow-based take global view. Optionally, memetic evolves pool solution candidates overall evaluate large set instances from application domains. With respect quality, outperforms all previously considered systems handle hypergraphs such hMETIS, PaToH, Mondriaan, or Zoltan. Somewhat surprisingly, some extend, even extends graph partitioners KaHIP when considering special case also faster these except PaToH represents different speed–quality
منابع مشابه
High Quality Hypergraph Partitioning via Max-Flow-Min-Cut Computations
In this thesis, we introduce a framework based on Max-Flow-Min-Cut computations for improving balanced k-way partitions of hypergraphs. Currently, variations of the FM heuristic [17] are used as local search algorithms in all state-of-the-art multilevel hypergraph partitioners. Such move-based heuristics have the disadvantage that they only incorporate local information about the problem struct...
متن کاملMemetic Multilevel Hypergraph Partitioning
Hypergraph partitioning has a wide range of important applications such as VLSI design or scientific computing. With focus on solution quality, we develop the first multilevel memetic algorithm to tackle the problem. Key components of our contribution are new effective multilevel recombination and mutation operations that provide a large amount of diversity. We perform a wide range of experimen...
متن کاملHypergraph Partitioning Techniques
Graph and hypergraph partitioning are important problems with applications in several areas including parallel processing, VLSI design automation, physical mapping of chromosomes and data classification. The problem is to divide the vertices of a hypergraph into k similar–sized blocks such that a given cost function defined over the hypergraph is optimized. The intent of this paper is to expose...
متن کاملEvolutionary Hypergraph Partitioning
The hypergraph partitioning problem has many applications like processor load balance or VLSI design. For most applications solution quality is crucial. Since the problem is NP-hard, heuristics and meta heuristics are used in practice to solve the problem. In this thesis we combine the commonly used multilevel heuristic with an evolutionary algorithm. Experimental results show that our new algo...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
ژورنال
عنوان ژورنال: ACM Journal of Experimental Algorithms
سال: 2022
ISSN: ['1084-6654']
DOI: https://doi.org/10.1145/3529090