arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2106.02206cs.LGcs.AIstat.ML
收藏

随机迭代图匹配

Stochastic Iterative Graph Matching

Tufts University(塔夫茨大学)

Linfeng Liu, Michael C. Hughes, Soha Hassoun, Li-Ping Liu

摘要:

最近的工作利用图神经网络来处理图匹配任务已经显示出有希望的结果。最近在学习离散分布方面的进展为学习图匹配模型提供了新的机会。在这项工作中,我们提出了一个新的模型,随机迭代图匹配(SIGMA),以解决图匹配问题。我们的模型定义了一个图对的匹配分布,因此模型可以探索各种可能的匹配。我们进一步介绍了一种新的多步匹配过程,它学习如何逐步完善一个图对的匹配结果。该模型还包括虚拟节点,以便该模型不必为没有对应关系的节点寻找匹配。我们通过可扩展的随机优化来拟合这个模型。我们在合成图形数据集以及生物化学和计算机视觉应用中进行了广泛的实验。在所有任务中,我们的结果表明,与最先进的模型相比,SIGMA可以显着改善图匹配结果。消融研究证实,我们的每个组件(随机训练,迭代匹配和虚拟节点)都提供了显着的改善。

英文摘要:

Recent works leveraging Graph Neural Networks to approach graph matching tasks have shown promising results. Recent progress in learning discrete distributions poses new opportunities for learning graph matching models. In this work, we propose a new model, Stochastic Iterative Graph MAtching (SIGMA), to address the graph matching problem. Our model defines a distribution of matchings for a graph pair so the model can explore a wide range of possible matchings. We further introduce a novel multi-step matching procedure, which learns how to refine a graph pair's matching results incrementally. The model also includes dummy nodes so that the model does not have to find matchings for nodes without correspondence. We fit this model to data via scalable stochastic optimization. We conduct extensive experiments across synthetic graph datasets as well as biochemistry and computer vision applications. Across all tasks, our results show that SIGMA can produce significantly improved graph matching results compared to state-of-the-art models. Ablation studies verify that each of our components (stochastic training, iterative matching, and dummy nodes) offers noticeable improvement.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑