Combinatorial Learning of Graph Edit Distance via Dynamic Embedding

Wang, Runzhong; Zhang, Tianqi; Yu, Tianshu; Yan, Junchi; Yang, Xiaokang

doi:10.1109/cvpr46437.2021.00520

Cited by 22 publications

(7 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The baselines encompass GNN approaches reported in the GED-CDA paper [41] and the SimGNN paper [16]. SimGNN [16], GraphSim [42], GMN [15], MPNGMN [43], GENN [44], and GED-CDA [41] are characterized by their focus on node-wise interactions. This results in a substantial computational burden due to their high complexity, which is at least O(|V A ||V B |).…”

Section: ) Graph Similarity Predictionmentioning

confidence: 99%

Co-Attention Graph Pooling for Efficient Pairwise Graph Interaction Learning

et al. 2023

View full text Add to dashboard Cite

Graph Neural Networks (GNNs) have proven to be effective in processing and learning from graph-structured data. However, previous works mainly focused on understanding single graph inputs while many real-world applications require pair-wise analysis for graph-structured data (e.g., scene graph matching, code searching, and drug-drug interaction prediction). To this end, recent works have shifted their focus to learning the interaction between pairs of graphs. Despite their improved performance, these works were still limited in that the interactions were considered at the node-level, resulting in high computational costs and suboptimal performance. To address this issue, we propose a novel and efficient graph-level approach for extracting interaction representations using co-attention in graph pooling. Our method, Co-Attention Graph Pooling (CAGPool), exhibits competitive performance relative to existing methods in both classification and regression tasks using real-world datasets, while maintaining lower computational complexity.

show abstract

Section: ) Graph Similarity Predictionmentioning

confidence: 99%

Co-Attention Graph Pooling for Efficient Pairwise Graph Interaction Learning

et al. 2023

View full text Add to dashboard Cite

show abstract

“…Computing GED is known NP-hard in general [37], therefore approximations are used. There are also attempts by deep graph model [19,33] and we adopt the same setting. In detail, triplet pairs are constructed by editing graph (substitute and remove) edges, which is an unsupervised model.…”

Section: Experiments Setupmentioning

confidence: 99%

DOTIN: Dropping Task-Irrelevant Nodes for GNNs

Zhang¹,

Zhu²,

Yan³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Scalability is an important consideration for deep graph neural networks. Inspired by the conventional pooling layers in CNNs, many recent graph learning approaches have introduced the pooling strategy to reduce the size of graphs for learning, such that the scalability and efficiency can be improved. However, these pooling-based methods are mainly tailored to a single graph-level task and pay more attention to local information, limiting their performance in multi-task settings which often require task-specific global information. In this paper, departure from these poolingbased efforts, we design a new approach called DOTIN (Dropping Task-Irrelevant Nodes) to reduce the size of graphs. Specifically, by introducing K learnable virtual nodes to represent the graph embeddings targeted to K different graph-level tasks, respectively, up to 90% raw nodes with low attentiveness with an attention model -a transformer in this paper, can be adaptively dropped without notable performance decreasing. Achieving almost the same accuracy, our method speeds up GAT about 50% on graph-level tasks including graph classification and graph edit distance (GED) with about 60% less memory, on D&D dataset. Code will be made public available in https://github.com/Sherrylone/DOTIN.

show abstract

“…There are efforts to learn graph-wise similarity via deep learning [3,32], which are regression models and ignores the combinatorial nature of graph similarity problems. [58] proposes neural-guided A* search, however, the learning is supervised. In this paper, we compare with PPO-Single by reimplementing [34] for GED.…”

Section: Case 2: Graph Edit Distancementioning

confidence: 99%

“…On the one hand, it is challenging to design a model with enough capacity with limited computational resources, and existing models are usually tailored for specific problems which require heavy trailand-error [24,55,57]. On the other hand, training such a heavy model requires either supervision from high-quality labels [29,55,58] which are infeasible to obtain for large-sized problems due to the NP-hard nature, or reinforcement learning (RL) [8,28,36,37] which might be unstable due to the challenges of large action space and sparse reward especially for large-sized problems [50]. Lower level: learning-free heuristic Figure 1: An overview of our bi-level hybrid MLCO solver.…”

Section: Introductionmentioning

confidence: 99%

“…ML modules are studied as sub-routines for traditional solvers, especially predicting branching strategies for branch-and-bound with either supervised learning [16,29] or reinforcement learning [49]. In [58], the heuristic routine in A* algorithm is replaced by graph neural network to solve graph edit distance. However, these methods are tailored for special problems, and our aim is to develop a more general framework, where the learning part and heuristic module are two peers alternatively performed.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A Bi-Level Framework for Learning to Solve Combinatorial Optimization on Graphs

Wang

Zou

Gan

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

Combinatorial Optimization (CO) has been a long-standing challenging research topic featured by its NP-hard nature. Traditionally such problems are approximately solved with heuristic algorithms which are usually fast but may sacrifice the solution quality. Currently, machine learning for combinatorial optimization (MLCO) has become a trending research topic, but most existing MLCO methods treat CO as a single-level optimization by directly learning the end-to-end solutions, which are hard to scale up and mostly limited by the capacity of ML models given the high complexity of CO. In this paper, we propose a hybrid approach to combine the best of the two worlds, in which a bi-level framework is developed with an upper-level learning method to optimize the graph (e.g. add, delete or modify edges in a graph), fused with a lower-level heuristic algorithm solving on the optimized graph. Such a bi-level approach simplifies the learning on the original hard CO and can effectively mitigate the demand for model capacity. The experiments and results on several popular CO problems like Directed Acyclic Graph scheduling, Graph Edit Distance and Hamiltonian Cycle Problem show its effectiveness over manually designed heuristics and single-level learning methods. * This work was done while the author was working as an intern at Ant Group.Preprint. Under review.

show abstract

Combinatorial Learning of Graph Edit Distance via Dynamic Embedding

Cited by 22 publications

References 31 publications

Co-Attention Graph Pooling for Efficient Pairwise Graph Interaction Learning

Co-Attention Graph Pooling for Efficient Pairwise Graph Interaction Learning

DOTIN: Dropping Task-Irrelevant Nodes for GNNs

A Bi-Level Framework for Learning to Solve Combinatorial Optimization on Graphs

Contact Info

Product

Resources

About