Since calculating the exact similarity between two graphs is usually an NP-hard problem, the tradeoff between precision and speed needs to be addressed. In this paper, a pooling-based graph neural network method is proposed, which effectively integrates the coarse-grained interaction features of the graph data and the fine-grained interaction features of the nodes between the subgraphs, and further reduce the computational cost while ensuring the accuracy. The experimental results show that the proposed method has good performance on real graph data sets. Compared with existing methods, the proposed method not only improves the accuracy but also improves the computational efficiency.
图分析的一类基本问题是计算图之间的相似度或距离。其中,图编辑距离(Graph edit distance,GED)和最大公共子图(Maximum common subgraph,MCS)是较常用的度量方法,但其精确计算需要较大的计算开销。为解决这一问题,一种简单有效的方法是利用GNN将输入图编码为图级嵌入,随后通过交互计算相似度。但这种方法存在缺陷,即图的差异可能存在于微小子结构中,而图级交互会忽视细粒度的交互特征。
KipfT N, WellingM. Semi-supervised classification with graph convolutional networks[J/OL].[2016-02-16].
[6]
HamiltonW, YingZ, LeskovecJ. Inductive representation learning on large graphs[J]. Advances in Neural Information Processing Systems, 2017, 30: 1024-1034.
[7]
VeličkovićP, CucurullG, CasanovaA, et al. Graph attention networks[[J/OL].[2017-06-16].
[8]
BrodyS, AlonU, YahavE. How attentive are graph attention networks?[J/OL].[2021-06-26].
[9]
YingZ, YouJ, MorrisC, et al. Hierarchical graph representation learning with differentiable pooling[J]. Advances in Neural Information Processing Systems, 2018, 31: 4800-4810.
[10]
GaoH Y, JiS W. Graph U-nets[C]∥Proceedings of the 36th International Conference on Machine Learning. New York: Curran Associates, Inc., International Machine Learning Society, 2019: 2083-2092.
[11]
LeeJ, LeeI, KangJ. Self-attention graph pooling[C]∥Proceedings of the 36th International Conference on Machine Learning. New York: Curran Associates, Inc., International Machine Learning Society, 2019: 3734-3743.
[12]
LuoY, McThrowM, AuW Y, et al. Automated data augmentations for graph classification[J/OL].[2022-08-11].
[13]
BaiY, DingH, BianS, et al. SimGNN: A neural network approach to fast graph similarity computation[C]∥Proceedings of the Twelfth ACM International Conference on Web Search and Data Mining. Yew York: ACM, 2019: 384-392.
[14]
BaiY, DingH, GuK, et al. Learning-based efficient graph similarity computation via multi-scale convolutional set matching[J]. AAAI, 2020, 34(4): 3219-3226.
[15]
KrizhevskyA, SutskeverI, HintonG E. Imagenet classification with deep convolutional neural networks[J]. Advances in Neural Information Processing Systems, 2012, 25: 1106-1114.
[16]
LingX, WuL, WangS,et al.Multilevel graph matching networks for deep graph similarity learning[J]. IEEE Transactions on Neural Networks and Learning Systems, 2023, 34: 799-813.
[17]
QinC, ZhaoH, WangL, et al. Slow learning and fast inference: Efficient graph similarity computation via knowledge distillation[J]. Advances in Neural Information Processing Systems, 2021, 34: 14110-14121.
[18]
RanjanR, GroverS, MedyaS, et al. GREED: a neural framework for learning graph distance functions[J]. Advances in Neural Information Processing Systems, 2022, 35: 22518-22530.
[19]
ZhuoW, TanG. Efficient graph similarity computation with alignment regularization[J]. Advances in Neural Information Processing Systems, 2022, 35: 30181-30193.
[20]
QureshiR J, RamelJ Y, CardotH. Graph based shapes representation and recognition[C]∥Graph-Based Representations in Pattern Recognition: 6th IAPR-TC-15 International Workshop. Berlin: Springer, 2007: 49-60.
[21]
BromleyJ, GuyonI, LeCunY, et al. Signature verification using a "siamese" time delay neural network[J]. Advances in Neural Information Processing Systems, 1993, 6: 737-744.
[22]
SocherR, ChenD, ManningC D, et al. Reasoning with neural tensor networks for knowledge base completion[J]. Advances in Neural Information Processing Systems, 2013, 26: 926-934.
[23]
WangX, DingX, TungA K H, et al. An efficient graph indexing method[C]∥Proceedings of the 28th IEEE International Conference on Data Engineering, Piscataway, USA, 2012: 210-221.
[24]
YanardagP, VishwanathanS V N. Deep graph kernels[C]∥Proceedings of the 21th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, New York,USA, 2015: 1365-1374.
[25]
NeuhausM, RiesenK, BunkeH. Fast suboptimal algorithms for the computation of graph edit distance[C]∥Structural, Syntactic, and Statistical Pattern Recognition, Berlin,Germany, 2006: 163-172.
[26]
KuhnH W. The Hungarian method for the assignment problem[J]. Naval Research Logistics Quarterly, 1955, 2(1/2): 83-97.
[27]
RiesenK, BunkeH. Approximate graph edit distance computation by means of bipartite graph matching[J]. Image and Vision Computing, 2009, 27(7): 950-959.
[28]
FankhauserS, RiesenK, BunkeH. Speeding up graph edit distance computation through fast bipartite matching[C]∥Graph-Based Representations in Pattern Recognition: The 8th IAPR-TC-15 International Workshop, Berlin,Germany, 2011: 102-111.
[29]
JonkerR, VolgenantT. A shortest augmenting path algorithm for dense and sparse linear assignment problems[J]. Computing, 1987, 38(4): 325-340.
[30]
DefferrardM, BressonX, VandergheynstP. Convolutional neural networks on graphs with fast localized spectral filtering[J]. Advances in Neural Information Processing Systems, 2016, 29: 3844-3852.