Impact of Memory Space Optimization Technique on Fast Network Motif Search Algorithm

Himanshu, -; Jain, Sarika

doi:10.1007/978-981-10-3770-2_52

Cited by 9 publications

(5 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Besides, some approaches cannot handle certain kinds of queries or input graphs. For example, SCMD (Wang et al 2012) and PATCOMP (Jain et al 2017) are not applicable to directed graphs, and PGD (Ahmed et al 2015) and RAGE (Marcus and Shavitt 2012) do not handle graphs with node or edge labels. Although Peregrine does not support directed edges or edge labels either, we run it on our datasets by ignoring edge directions/labels if any, and focus on its time cost only.…”

Section: Methodsmentioning

confidence: 99%

Learning to Count Isomorphisms with Graph Neural Networks

Liu

Fang

et al. 2023

AAAI

View full text Add to dashboard Cite

Subgraph isomorphism counting is an important problem on graphs, as many graph-based tasks exploit recurring subgraph patterns. Classical methods usually boil down to a backtracking framework that needs to navigate a huge search space with prohibitive computational cost. Some recent studies resort to graph neural networks (GNNs) to learn a low-dimensional representation for both the query and input graphs, in order to predict the number of subgraph isomorphisms on the input graph. However, typical GNNs employ a node-centric message passing scheme that receives and aggregates messages on nodes, which is inadequate in complex structure matching for isomorphism counting. Moreover, on an input graph, the space of possible query graphs is enormous, and different parts of the input graph will be triggered to match different queries. Thus, expecting a fixed representation of the input graph to match diversely structured query graphs is unrealistic. In this paper, we propose a novel GNN called Count-GNN for subgraph isomorphism counting, to deal with the above challenges. At the edge level, given that an edge is an atomic unit of encoding graph structures, we propose an edge-centric message passing scheme, where messages on edges are propagated and aggregated based on the edge adjacency to preserve fine-grained structural information. At the graph level, we modulate the input graph representation conditioned on the query, so that the input graph can be adapted to each query individually to improve their matching. Finally, we conduct extensive experiments on a number of benchmark datasets to demonstrate the superior performance of Count-GNN.

show abstract

Section: Methodsmentioning

confidence: 99%

Learning to Count Isomorphisms with Graph Neural Networks

Liu

Fang

et al. 2023

AAAI

View full text Add to dashboard Cite

show abstract

“…Both Quatexelero and FaSE have potential memory issues, since there may be several nodes representing the same graph, which is not a problem for Gtrie since it only stores one copy of each possible graph. To address this, Himamshu and Jain [61] proposed Patcomp. Their method compresses the quaternary tree using a technique similar to a radix tree, however, their method is 2 to 3 times slower and only saves around 10% of the memory usage.…”

Section: Encapsulation Methodsmentioning

confidence: 99%

A Survey on Subgraph Counting: Concepts, Algorithms and Applications to Network Motifs and Graphlets

Ribeiro,

Paredes,

Silva

et al. 2019

Preprint

View full text Add to dashboard Cite

Computing subgraph frequencies is a fundamental task that lies at the core of several network analysis methodologies, such as network motifs and graphlet-based metrics, which have been widely used to categorize and compare networks from multiple domains. Counting subgraphs is however computationally very expensive and there has been a large body of work on efficient algorithms and strategies to make subgraph counting feasible for larger subgraphs and networks.This survey aims precisely to provide a comprehensive overview of the existing methods for subgraph counting. Our main contribution is a general and structured review of existing algorithms, classifying them on a set of key characteristics, highlighting their main similarities and differences. We identify and describe the main conceptual approaches, giving insight on their advantages and limitations, and provide pointers to existing implementations. We initially focus on exact sequential algorithms, but we also do a thorough survey on approximate methodologies (with a trade-off between accuracy and execution time) and parallel strategies (that need to deal with an unbalanced search space).

show abstract

“…All the worker processes during the enumeration and census maintain their individual copies of PATRICIA trie and ZeroOne binary tree. Details of PATRICIA trie and ZeroOne tree data structure and their application have been illustrated in sequential algorithm based network motif search process (Himanshu et al 2017). When a worker receives the message with 'FINISH' tag, it sends three lists viz.…”

Section: Methodsmentioning

confidence: 99%

“…It could be ascertained that as we distribute the search tree branches to the processing cores of distributed-memory cluster, each core would require to maintain its own individual copy of data structure to store the subgraph isomorphism details in memory so that number of calls to 'nauty' could be reduced for identification of subgraph isomorphism each time a new subgraph is enumerated. Further, it is observed that as the query subgraph size increases, size of g-trie (Ribeiro and Silva 2009) and quaternary tree (Khakabimamaghani et al 2013) increases much higher than PATRICIA trie (Himanshu et al 2017). So g-trie and quaternary tree based algorithms run the risk of becoming infeasible on distributed-memory cluster if parallel work units are assigned to more number of cores per processor, as the motif size becomes large.…”

mentioning

confidence: 99%

Parallel network motif search using message passing approach for biological complex networks

Himanshu

Bandyopadhyay

Jain

2017

Indian J of Anim Sci

View full text Add to dashboard Cite

Study of complex biological networks is essential for understanding their functional characteristics. Network motifs have functional significance in biological networks as they represent building blocks of these networks. This study evaluates master-worker parallelization approach on sequential PATRICIA trie based fast network motif search algorithm for distributed memory model based High Performance Clusters (HPCs). Proposed algorithm uses PATRICIA trie for data compression during census of subgraphs based upon ESU algorithm. Parallel implementation was done using MPI and C language. We applied proposed parallel algorithm to three real networks viz. networks of metabolic pathway of E.coli, electronic and social networks. PATRICIA based parallel approach was able to achieve speedup of 50.75, 49.37, 38.07 as analysed on 101 cores on networks of metabolic pathway of E.coli, electronic and social networks respectively for large motifs of size 9 for E.coli, social and 10 for electronic networks over the PATRICIA trie based sequential algorithm.

show abstract

Impact of Memory Space Optimization Technique on Fast Network Motif Search Algorithm

Cited by 9 publications

References 7 publications

Learning to Count Isomorphisms with Graph Neural Networks

Learning to Count Isomorphisms with Graph Neural Networks

A Survey on Subgraph Counting: Concepts, Algorithms and Applications to Network Motifs and Graphlets

Parallel network motif search using message passing approach for biological complex networks

Contact Info

Product

Resources

About