arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2109.14218cs.LG
收藏

因子图的等变神经网络

Equivariant Neural Network for Factor Graphs

Stanford University(斯坦福大学);Dexterity(德克斯特公司);Cornell University(康奈尔大学)

Fan-Yun Sun, Jonathan Kuck, Hao Tang, Stefano Ermon

摘要:

在因子图数据结构中使用的若干索引可以被置换而不改变潜在的概率分布。在因子图上执行推理的算法理想地应该对节点的全局索引、因子内的变量排序和变量赋值排序的排列是等变的或不变的。然而,现有的基于神经网络的推理过程未能利用这种归纳偏差。在本文中,我们精确地描述了这些同构的因素图的性质,并提出了两个推理模型:因子-等变神经信念传播(FE-NBP)和因子-等变图神经网络(FE-GNN)。FE-NBP是一个神经网络,它概括了BP并尊重因子图的上述每个属性,而FE-GNN是一个表达性GNN模型,它放松了同构属性,有利于更大的表达性。从经验上讲,我们在真实世界和合成数据集上证明,对于边缘推理和MAP推理,FE-NBP和FE-GNN共同涵盖了一系列样本复杂性机制:FE-NBP在小数据集上实现了最先进的性能,而FE-GNN在大数据集上实现了最先进的性能。

英文摘要:

Several indices used in a factor graph data structure can be permuted without changing the underlying probability distribution. An algorithm that performs inference on a factor graph should ideally be equivariant or invariant to permutations of global indices of nodes, variable orderings within a factor, and variable assignment orderings. However, existing neural network-based inference procedures fail to take advantage of this inductive bias. In this paper, we precisely characterize these isomorphic properties of factor graphs and propose two inference models: Factor-Equivariant Neural Belief Propagation (FE-NBP) and Factor-Equivariant Graph Neural Networks (FE-GNN). FE-NBP is a neural network that generalizes BP and respects each of the above properties of factor graphs while FE-GNN is an expressive GNN model that relaxes an isomorphic property in favor of greater expressivity. Empirically, we demonstrate on both real-world and synthetic datasets, for both marginal inference and MAP inference, that FE-NBP and FE-GNN together cover a range of sample complexity regimes: FE-NBP achieves state-of-the-art performance on small datasets while FE-GNN achieves state-of-the-art performance on large datasets.

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。

↑