An official website of the United States government
Transparency and accountability have become major concerns for black-box machine learning (ML) models. Proper explanations for the model behavior increase model transparency and help researchers develop more accountable models. Graph neural networks (GNN) have recently shown superior performance in many graph ML problems than traditional methods, and explaining them has attracted increased interest. However, GNN explanation for link prediction (LP) is lacking in the literature. LP is an essential GNN task and corresponds to web applications like recommendation and sponsored search on web. Given existing GNN explanation methods only address node/graph-level tasks, we propose Path-based GNN Explanation for heterogeneous Link prediction (PaGE-Link) that generates explanations with connection interpretability, enjoys model scalability, and handles graph heterogeneity. Qualitatively, PaGE-Link can generate explanations as paths connecting a node pair, which naturally captures connections between the two nodes and easily transfer to human-interpretable explanations. Quantitatively, explanations generated by PaGE-Link improve AUC for recommendation on citation and user-item graphs by 9 - 35% and are chosen as better by 78.79% of responses in human evaluation.
more »
« less
On analyzing graphs with motif-paths
Path-based solutions have been shown to be useful for various graph analysis tasks, such as link prediction and graph clustering. However, they are no longer adequate for handling complex and gigantic graphs. Recently, motif-based analysis has attracted a lot of attention. A motif, or a small graph with a few nodes, is often considered as a fundamental unit of a graph. Motif-based analysis captures high-order structure between nodes, and performs better than traditional "edge-based" solutions. In this paper, we study motif-path , which is conceptually a concatenation of one or more motif instances. We examine how motif-paths can be used in three path-based mining tasks, namely link prediction, local graph clustering and node ranking. We further address the situation when two graph nodes are not connected through a motif-path, and develop a novel defragmentation method to enhance it. Experimental results on real graph datasets demonstrate the use of motif-paths and defragmentation techniques improves graph analysis effectiveness.
more »
« less
- Award ID(s):
- 1633755
- PAR ID:
- 10427941
- Date Published:
- Journal Name:
- Proceedings of the VLDB Endowment
- Volume:
- 14
- Issue:
- 6
- ISSN:
- 2150-8097
- Page Range / eLocation ID:
- 1111 to 1123
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
null (Ed.)This work proposes a new unsupervised (or self-supervised) node representation learning method that aims to leverage the coarse-grain information that is available in most graphs. This extends previous attempts that only leverage fine-grain information (similarities within local neighborhoods) or global graph information (similarities across all nodes). Intuitively, the proposed method identifies nodes that belong to the same clusters and maximizes their mutual information. Thus, coarse-grain (cluster-level) similarities that are shared between nodes are preserved in their representations. The core components of the proposed method are (i) a jointly optimized clustering of nodes during learning and (ii) an Infomax objective term that preserves the mutual information among nodes of the same clusters. Our method is able to outperform competing state-of-art methods in various downstream tasks, such as node classification, link prediction, and node clustering. Experiments show that the average gain is between 0.2% and 6.1%, over the best competing approach, over all tasks. Our code is publicly available at: https://github.com/cmavro/Graph-InfoClust-GIC.more » « less
-
null (Ed.)Message passing Graph Neural Networks (GNNs) provide a powerful modeling framework for relational data. However, the expressive power of existing GNNs is upper-bounded by the 1-Weisfeiler-Lehman (1-WL) graph isomorphism test, which means GNNs that are not able to predict node clustering coefficients and shortest path distances, and cannot differentiate between different d regular graphs. Here we develop a class of message passing GNNs, named Identity-aware Graph Neural Networks (ID-GNNs), with greater expressive power than the 1-WL test. ID-GNN offers a minimal but powerful solution to limitations of existing GNNs. ID-GNN extends existing GNN architectures by inductively considering nodes’ identities during message passing. To embed a given node, IDGNN first extracts the ego network centered at the node, then conducts rounds of heterogeneous message passing, where different sets of parameters are applied to the center node than to other surrounding nodes in the ego network. We further propose a simplified but faster version of ID-GNN that injects node identity information as augmented node features. Altogether, both versions of ID GNN represent general extensions of message passing GNNs, where experiments show that transforming existing GNNs to ID-GNNs yields on average 40% accuracy improvement on challenging node, edge, and graph property prediction tasks; 3% accuracy improvement on node and graph classification benchmarks; and 15% ROC AUC improvement on real-world link prediction tasks. Additionally, ID-GNNs demonstrate improved or comparable performance over other task-specific graph networks.more » « less
-
null (Ed.)Message passing Graph Neural Networks (GNNs) provide a powerful modeling framework for relational data. However, the expressive power of existing GNNs is upper-bounded by the 1-Weisfeiler-Lehman (1-WL) graph isomorphism test, which means GNNs that are not able to predict node clustering coefficients and shortest path distances, and cannot differentiate between different d-regular graphs. Here we develop a class of message passing GNNs, named Identity-aware Graph Neural Networks (ID-GNNs), with greater expressive power than the 1-WL test. ID-GNN offers a minimal but powerful solution to limitations of existing GNNs. ID-GNN extends existing GNN architectures by inductively considering nodes’ identities during message passing. To embed a given node, IDGNN first extracts the ego network centered at the node, then conducts rounds of heterogeneous message passing, where different sets of parameters are applied to the center node than to other surrounding nodes in the ego network. We further propose a simplified but faster version of ID-GNN that injects node identity information as augmented node features. Altogether, both versions of ID-GNN represent general extensions of message passing GNNs, where experiments show that transforming existing GNNs to ID-GNNs yields on average 40% accuracy improvement on challenging node, edge, and graph property prediction tasks; 3% accuracy improvement on node and graph classification benchmarks; and 15% ROC AUC improvement on real-world link prediction tasks. Additionally, ID-GNNs demonstrate improved or comparable performance over other task-specific graph networks.more » « less
-
In this paper, we propose a novel representation learning framework, namely HIN2Vec, for heterogeneous information networks (HINs). The core of the proposed framework is a neural network model, also called HIN2Vec, designed to capture the rich semantics embedded in HINs by exploiting different types of relationships among nodes. Given a set of relationships specified in forms of meta-paths in an HIN, HIN2Vec carries out multiple prediction training tasks jointly based on a target set of relationships to learn latent vectors of nodes and meta-paths in the HIN. In addition to model design, several issues unique to HIN2Vec, including regularization of meta-path vectors, node type selection in negative sampling, and cycles in random walks, are examined. To validate our ideas, we learn latent vectors of nodes using four large-scale real HIN datasets, including Blogcatalog, Yelp, DBLP and U.S. Patents, and use them as features for multi-label node classification and link prediction applications on those networks. Empirical results show that HIN2Vec soundly outperforms the state-of-the-art representation learning models for network data, including DeepWalk, LINE, node2vec, PTE, HINE and ESim, by 6.6% to 23.8% ofmicro-f1 in multi-label node classification and 5% to 70.8% of MAP in link prediction.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript
- Journal Article:
- https://doi.org/10.14778/3447689.3447714
