scholarly journals Graph Transformer for Graph-to-Sequence Learning

2020 ◽  
Vol 34 (05) ◽  
pp. 7464-7471
Author(s):  
Deng Cai ◽  
Wai Lam
Keyword(s):  
Neural Networks ◽  
Information Exchange ◽  
State Of The Art ◽  
Structural Information ◽  
Representation Learning ◽  
Graph Representation ◽  
Text Generation ◽  
Proposed Model ◽  
Graph Neural Networks ◽  
Meaning Representation

The dominant graph-to-sequence transduction models employ graph neural networks for graph representation learning, where the structural information is reflected by the receptive field of neurons. Unlike graph neural networks that restrict the information exchange between immediate neighborhood, we propose a new model, known as Graph Transformer, that uses explicit relation encoding and allows direct communication between two distant nodes. It provides a more efficient way for global graph structure modeling. Experiments on the applications of text generation from Abstract Meaning Representation (AMR) and syntax-based neural machine translation show the superiority of our proposed model. Specifically, our model achieves 27.4 BLEU on LDC2015E86 and 29.7 BLEU on LDC2017T10 for AMR-to-text generation, outperforming the state-of-the-art results by up to 2.2 points. On the syntax-based translation tasks, our model establishes new single-model state-of-the-art BLEU scores, 21.3 for English-to-German and 14.1 for English-to-Czech, improving over the existing best results, including ensembles, by over 1 BLEU.

scholarly journals Multi-hop Attention Graph Neural Networks

2021 ◽  
Author(s):  
Guangtao Wang ◽  
Rex Ying ◽  
Jing Huang ◽  
Jure Leskovec
Keyword(s):  
Neural Networks ◽  
Large Scale ◽  
Performance Metrics ◽  
State Of The Art ◽  
Structural Information ◽  
Representation Learning ◽  
Attention Mechanism ◽  
Graph Representation ◽  
Knowledge Graph ◽  
Graph Neural Networks

Self-attention mechanism in graph neural networks (GNNs) led to state-of-the-art performance on many graph representation learning tasks. Currently, at every layer, attention is computed between connected pairs of nodes and depends solely on the representation of the two nodes. However, such attention mechanism does not account for nodes that are not directly connected but provide important network context. Here we propose Multi-hop Attention Graph Neural Network (MAGNA), a principled way to incorporate multi-hop context information into every layer of attention computation. MAGNA diffuses the attention scores across the network, which increases the receptive field for every layer of the GNN. Unlike previous approaches, MAGNA uses a diffusion prior on attention values, to efficiently account for all paths between the pair of disconnected nodes. We demonstrate in theory and experiments that MAGNA captures large-scale structural information in every layer, and has a low-pass effect that eliminates noisy high-frequency information from graph data. Experimental results on node classification as well as the knowledge graph completion benchmarks show that MAGNA achieves state-of-the-art results: MAGNA achieves up to 5.7% relative error reduction over the previous state-of-the-art on Cora, Citeseer, and Pubmed. MAGNA also obtains the best performance on a large-scale Open Graph Benchmark dataset. On knowledge graph completion MAGNA advances state-of-the-art on WN18RR and FB15k-237 across four different performance metrics.

scholarly journals UniGNN: a Unified Framework for Graph and Hypergraph Neural Networks

2021 ◽  
Author(s):  
Jing Huang ◽  
Jie Yang
Keyword(s):  
Neural Networks ◽  
Message Passing ◽  
State Of The Art ◽  
Representation Learning ◽  
Graph Representation ◽  
Challenging Problem ◽  
Unified Framework ◽  
Real World Datasets ◽  
Graph Neural Networks ◽  
Research Domains

Hypergraph, an expressive structure with flexibility to model the higher-order correlations among entities, has recently attracted increasing attention from various research domains. Despite the success of Graph Neural Networks (GNNs) for graph representation learning, how to adapt the powerful GNN-variants directly into hypergraphs remains a challenging problem. In this paper, we propose UniGNN, a unified framework for interpreting the message passing process in graph and hypergraph neural networks, which can generalize general GNN models into hypergraphs. In this framework, meticulously-designed architectures aiming to deepen GNNs can also be incorporated into hypergraphs with the least effort. Extensive experiments have been conducted to demonstrate the effectiveness of UniGNN on multiple real-world datasets, which outperform the state-of-the-art approaches with a large margin. Especially for the DBLP dataset, we increase the accuracy from 77.4% to 88.8% in the semi-supervised hypernode classification task. We further prove that the proposed message-passing based UniGNN models are at most as powerful as the 1-dimensional Generalized Weisfeiler-Leman (1-GWL) algorithm in terms of distinguishing non-isomorphic hypergraphs. Our code is available at https://github.com/OneForward/UniGNN.

scholarly journals GSSNN: Graph Smoothing Splines Neural Networks

2020 ◽  
Vol 34 (04) ◽  
pp. 7007-7014
Author(s):  
Shichao Zhu ◽  
Lewei Zhou ◽  
Shirui Pan ◽  
Chuan Zhou ◽  
Guiying Yan ◽  
...  
Keyword(s):  
Neural Networks ◽  
State Of The Art ◽  
Representation Learning ◽  
Smoothing Splines ◽  
Graph Representation ◽  
Graph Classification ◽  
Topological Information ◽  
Graph Representations ◽  
The Arts ◽  
Graph Neural Networks

Graph Neural Networks (GNNs) have achieved state-of-the-art performance in many graph data analysis tasks. However, they still suffer from two limitations for graph representation learning. First, they exploit non-smoothing node features which may result in suboptimal embedding and degenerated performance for graph classification. Second, they only exploit neighbor information but ignore global topological knowledge. Aiming to overcome these limitations simultaneously, in this paper, we propose a novel, flexible, and end-to-end framework, Graph Smoothing Splines Neural Networks (GSSNN), for graph classification. By exploiting the smoothing splines, which are widely used to learn smoothing fitting function in regression, we develop an effective feature smoothing and enhancement module Scaled Smoothing Splines (S3) to learn graph embedding. To integrate global topological information, we design a novel scoring module, which exploits closeness, degree, as well as self-attention values, to select important node features as knots for smoothing splines. These knots can be potentially used for interpreting classification results. In extensive experiments on biological and social datasets, we demonstrate that our model achieves state-of-the-arts and GSSNN is superior in learning more robust graph representations. Furthermore, we show that S3 module is easily plugged into existing GNNs to improve their performance.

scholarly journals An Attention-Based Graph Neural Network for Heterogeneous Structural Learning

2020 ◽  
Vol 34 (04) ◽  
pp. 4132-4139
Author(s):  
Huiting Hong ◽  
Hantao Guo ◽  
Yucheng Lin ◽  
Xiaoqing Yang ◽  
Zang Li ◽  
...  
Keyword(s):  
Neural Network ◽  
Structural Information ◽  
Representation Learning ◽  
Graph Representation ◽  
Heterogeneous Information ◽  
Domain Experts ◽  
Proposed Model ◽  
Meta Path ◽  
Low Dimensional ◽  
Public Datasets

In this paper, we focus on graph representation learning of heterogeneous information network (HIN), in which various types of vertices are connected by various types of relations. Most of the existing methods conducted on HIN revise homogeneous graph embedding models via meta-paths to learn low-dimensional vector space of HIN. In this paper, we propose a novel Heterogeneous Graph Structural Attention Neural Network (HetSANN) to directly encode structural information of HIN without meta-path and achieve more informative representations. With this method, domain experts will not be needed to design meta-path schemes and the heterogeneous information can be processed automatically by our proposed model. Specifically, we implicitly represent heterogeneous information using the following two methods: 1) we model the transformation between heterogeneous vertices through a projection in low-dimensional entity spaces; 2) afterwards, we apply the graph neural network to aggregate multi-relational information of projected neighborhood by means of attention mechanism. We also present three extensions of HetSANN, i.e., voices-sharing product attention for the pairwise relationships in HIN, cycle-consistency loss to retain the transformation between heterogeneous entity spaces, and multi-task learning with full use of information. The experiments conducted on three public datasets demonstrate that our proposed models achieve significant and consistent improvements compared to state-of-the-art solutions.

scholarly journals Pairwise Half-graph Discrimination: A Simple Graph-level Self-supervised Strategy for Pre-training Graph Neural Networks

2021 ◽  
Author(s):  
Pengyong Li ◽  
Jun Wang ◽  
Ziliang Li ◽  
Yixuan Qiao ◽  
Xianggen Liu ◽  
...  
Keyword(s):  
Neural Networks ◽  
Supervised Learning ◽  
Learning Strategy ◽  
Binary Classification ◽  
Representation Learning ◽  
Graph Representation ◽  
Superior Performance ◽  
Graph Classification ◽  
Training Strategy ◽  
Graph Neural Networks

Self-supervised learning has gradually emerged as a powerful technique for graph representation learning. However, transferable, generalizable, and robust representation learning on graph data still remains a challenge for pre-training graph neural networks. In this paper, we propose a simple and effective self-supervised pre-training strategy, named Pairwise Half-graph Discrimination (PHD), that explicitly pre-trains a graph neural network at graph-level. PHD is designed as a simple binary classification task to discriminate whether two half-graphs come from the same source. Experiments demonstrate that the PHD is an effective pre-training strategy that offers comparable or superior performance on 13 graph classification tasks compared with state-of-the-art strategies, and achieves notable improvements when combined with node-level strategies. Moreover, the visualization of learned representation revealed that PHD strategy indeed empowers the model to learn graph-level knowledge like the molecular scaffold. These results have established PHD as a powerful and effective self-supervised learning strategy in graph-level representation learning.

scholarly journals Molecular Graph Contrastive Learning with Parameterized Explainable Augmentations

2021 ◽  
Author(s):  
Yingheng Wang ◽  
Yaosen Min ◽  
Erzhuo Shao ◽  
Ji Wu
Keyword(s):  
Neural Networks ◽  
Deep Neural Networks ◽  
Structural Information ◽  
Molecular Graph ◽  
Representation Learning ◽  
Graph Representation ◽  
Input Graph ◽  
Recent Success ◽  
Real World Datasets ◽  
Comparative Results

ABSTRACTLearning generalizable, transferable, and robust representations for molecule data has always been a challenge. The recent success of contrastive learning (CL) for self-supervised graph representation learning provides a novel perspective to learn molecule representations. The most prevailing graph CL framework is to maximize the agreement of representations in different augmented graph views. However, existing graph CL frameworks usually adopt stochastic augmentations or schemes according to pre-defined rules on the input graph to obtain different graph views in various scales (e.g. node, edge, and subgraph), which may destroy topological semantemes and domain prior in molecule data, leading to suboptimal performance. Therefore, designing parameterized, learnable, and explainable augmentation is quite necessary for molecular graph contrastive learning. A well-designed parameterized augmentation scheme can preserve chemically meaningful structural information and intrinsically essential attributes for molecule graphs, which helps to learn representations that are insensitive to perturbation on unimportant atoms and bonds. In this paper, we propose a novel Molecular Graph Contrastive Learning with Parameterized Explainable Augmentations, MolCLE for brevity, that self-adaptively incorporates chemically significative information from both topological and semantic aspects of molecular graphs. Specifically, we apply deep neural networks to parameterize the augmentation process for both the molecular graph topology and atom attributes, to highlight contributive molecular substructures and recognize underlying chemical semantemes. Comprehensive experiments on a variety of real-world datasets demonstrate that our proposed method consistently outperforms compared baselines, which verifies the effectiveness of the proposed framework. Detailedly, our self-supervised MolCLE model surpasses many supervised counterparts, and meanwhile only uses hundreds of thousands of parameters to achieve comparative results against the state-of-the-art baseline, which has tens of millions of parameters. We also provide detailed case studies to validate the explainability of augmented graph views.CCS CONCEPTS• Mathematics of computing → Graph algorithms; • Applied computing → Bioinformatics; • Computing methodologies → Neural networks; Unsupervised learning.

Unsupervised Structural Graph Node Representation Learning

2020 ◽  
Author(s):  
Mikel Joaristi
Keyword(s):  
Real World ◽  
State Of The Art ◽  
Structural Information ◽  
Representation Learning ◽  
Graph Representation ◽  
Learning Methods ◽  
Structural Graph ◽  
Connectivity Information ◽  
Latent Space ◽  
Previous State

Unsupervised Graph Representation Learning methods learn a numerical representation of the nodes in a graph. The generated representations encode meaningful information about the nodes' properties, making them a powerful tool for tasks in many areas of study, such as social sciences, biology or communication networks. These methods are particularly interesting because they facilitate the direct use of standard Machine Learning models on graphs. Graph representation learning methods can be divided into two main categories depending on the information they encode, methods preserving the nodes connectivity information, and methods preserving nodes' structural information. Connectivity-based methods focus on encoding relationships between nodes, with neighboring nodes being closer together in the resulting latent space. On the other hand, structure-based methods generate a latent space where nodes serving a similar structural function in the network are encoded close to each other, independently of them being connected or even close to each other in the graph. While there are a lot of works that focus on preserving nodes' connectivity information, only a few works study the problem of encoding nodes' structure, specially in an unsupervised way. In this dissertation, we demonstrate that properly encoding nodes' structural information is fundamental for many real-world applications, as it can be leveraged to successfully solve many tasks where connectivity-based methods fail. One concrete example is presented first. In this example, the task consists of detecting malicious entities in a real-world financial network. We show that to solve this problem, connectivity information is not enough and show how leveraging structural information provides considerable performance improvements. This particular example pinpoints the need for further research on the area of structural graph representation learning, together with the limitations of the previous state-of-the-art. We use the acquired knowledge as a starting point and inspiration for the research and development of three independent unsupervised structural graph representation learning methods: Structural Iterative Representation learning approach for Graph Nodes (SIR-GN), Structural Iterative Lexicographic Autoencoded Node Representation (SILA), and Sparse Structural Node Representation (SparseStruct). We show how each of our methods tackles specific limitations on the previous state-of-the-art on structural graph representation learning such as scalability, representation meaning, and lack of formal proof that guarantees the preservation of structural properties. We provide an extensive experimental section where we compare our three proposed methods to the current state-of-the-art on both connectivity-based and structure-based representation learning methods. Finally, in this dissertation, we look at extensions of the basic structural graph representation learning problem. We study the problem of temporal structural graph representation. We also provide a method for representation explainability.

scholarly journals TSI-GNN: Extending Graph Neural Networks to Handle Missing Data in Temporal Settings

2021 ◽  
Vol 4 ◽  
Author(s):  
David Gordon ◽  
Panayiotis Petousis ◽  
Henry Zheng ◽  
Davina Zamanzadeh ◽  
Alex A.T. Bui
Keyword(s):  
Neural Networks ◽  
Missing Data ◽  
Bipartite Graph ◽  
Message Passing ◽  
Representation Learning ◽  
Temporal Information ◽  
Graph Representation ◽  
Sequence Information ◽  
Novel Approach ◽  
Graph Neural Networks

We present a novel approach for imputing missing data that incorporates temporal information into bipartite graphs through an extension of graph representation learning. Missing data is abundant in several domains, particularly when observations are made over time. Most imputation methods make strong assumptions about the distribution of the data. While novel methods may relax some assumptions, they may not consider temporality. Moreover, when such methods are extended to handle time, they may not generalize without retraining. We propose using a joint bipartite graph approach to incorporate temporal sequence information. Specifically, the observation nodes and edges with temporal information are used in message passing to learn node and edge embeddings and to inform the imputation task. Our proposed method, temporal setting imputation using graph neural networks (TSI-GNN), captures sequence information that can then be used within an aggregation function of a graph neural network. To the best of our knowledge, this is the first effort to use a joint bipartite graph approach that captures sequence information to handle missing data. We use several benchmark datasets to test the performance of our method against a variety of conditions, comparing to both classic and contemporary methods. We further provide insight to manage the size of the generated TSI-GNN model. Through our analysis we show that incorporating temporal information into a bipartite graph improves the representation at the 30% and 60% missing rate, specifically when using a nonlinear model for downstream prediction tasks in regularly sampled datasets and is competitive with existing temporal methods under different scenarios.

scholarly journals Node-wise Localization of Graph Neural Networks

2021 ◽  
Author(s):  
Zemin Liu ◽  
Yuan Fang ◽  
Chenghao Liu ◽  
Steven C.H. Hoi
Keyword(s):  
Neural Networks ◽  
State Of The Art ◽  
Representation Learning ◽  
Global Model ◽  
Local Context ◽  
Local Contexts ◽  
Unique Model ◽  
Graph Neural Networks ◽  
Global And Local ◽  
Different Parts

Graph neural networks (GNNs) emerge as a powerful family of representation learning models on graphs. To derive node representations, they utilize a global model that recursively aggregates information from the neighboring nodes. However, different nodes reside at different parts of the graph in different local contexts, making their distributions vary across the graph. Ideally, how a node receives its neighborhood information should be a function of its local context, to diverge from the global GNN model shared by all nodes. To utilize node locality without overfitting, we propose a node-wise localization of GNNs by accounting for both global and local aspects of the graph. Globally, all nodes on the graph depend on an underlying global GNN to encode the general patterns across the graph; locally, each node is localized into a unique model as a function of the global model and its local context. Finally, we conduct extensive experiments on four benchmark graphs, and consistently obtain promising performance surpassing the state-of-the-art GNNs.

scholarly journals Graph Entropy Guided Node Embedding Dimension Selection for Graph Neural Networks

2021 ◽  
Author(s):  
Gongxu Luo ◽  
Jianxin Li ◽  
Hao Peng ◽  
Carl Yang ◽  
Lichao Sun ◽  
...  
Keyword(s):  
Neural Networks ◽  
Fundamental Problem ◽  
Representation Learning ◽  
Graph Representation ◽  
Basic Unit ◽  
Order Structure ◽  
Great Success ◽  
Embedding Dimension ◽  
Minimum Entropy ◽  
Graph Neural Networks

Graph representation learning has achieved great success in many areas, including e-commerce, chemistry, biology, etc. However, the fundamental problem of choosing the appropriate dimension of node embedding for a given graph still remains unsolved. The commonly used strategies for Node Embedding Dimension Selection (NEDS) based on grid search or empirical knowledge suffer from heavy computation and poor model performance. In this paper, we revisit NEDS from the perspective of minimum entropy principle. Subsequently, we propose a novel Minimum Graph Entropy (MinGE) algorithm for NEDS with graph data. To be specific, MinGE considers both feature entropy and structure entropy on graphs, which are carefully designed according to the characteristics of the rich information in them. The feature entropy, which assumes the embeddings of adjacent nodes to be more similar, connects node features and link topology on graphs. The structure entropy takes the normalized degree as basic unit to further measure the higher-order structure of graphs. Based on them, we design MinGE to directly calculate the ideal node embedding dimension for any graph. Finally, comprehensive experiments with popular Graph Neural Networks (GNNs) on benchmark datasets demonstrate the effectiveness and generalizability of our proposed MinGE.

Sign in / Sign up

Export Citation Format

Share Document