Graph Attention Network Research Articles

LSTM-based graph attention network for vehicle trajectory prediction

Vehicle Trajectory Prediction (VTP) is one of the key technologies for autonomous driving, which can improve the safety and collaboration of the autonomous driving system. The interaction behavior among vehicles in reality has an impact on VTP. However, many methods ignore the interaction among vehicles, which results in limited accuracy of prediction results. Therefore, we propose a Long Short-Term Memory (LSTM)-based Graph Attention Network (GAT) method for VTP, which encodes vehicle trajectory information with LSTM networks and represents vehicle interactions with GAT. Firstly, in order to capture the temporal relationship between positions and consider their influence, we use LSTM model to encode the position data. Meanwhile, to comprehensively model vehicle motion and use multidimensional feature representation, we employ another LSTM model to encode the motion data, including position, velocity and acceleration. Secondly, to learn distinct feature representation, we use one GAT module to process the LSTM position encoding features for capturing spatial relationships of position information. Another GAT module is employed to process the LSTM motion encoding features for fully considering multidimensional motion dynamics and spatial–temporal dependencies. Finally, the LSTM decoder receives all features and predicts the vehicle trajectory. The experimental results show that the proposed method demonstrates superior predictive performance by using the Next Generation Simulation (NGSIM) dataset.

An IIoT Temporal Data Anomaly Detection Method Combining Transformer and Adversarial Training

The existing Industrial Internet of Things (IIoT) temporal data analysis methods often suffer from issues such as information loss, difficulty balancing spatial and temporal features, and being affected by training data noise, which can lead to varying degrees of reduced model accuracy. Therefore, a new anomaly detection method was proposed, which integrated Transformer and adversarial training. Firstly, a bidirectional spatiotemporal feature extraction module was constructed by combining Graph Attention Networks (GAT) and Bidirectional Gated Recurrent Unit (BiGRU), which can simultaneously extract spatial and temporal features. Then, by combining multi-scale convolution with Long Short-Term Memory (LSTM), multi-scale contextual information was captured. Finally, an improved Transformer was used to fuse multi-dimensional features, combined with an adversarial-trained variational autoencoder to calculate the anomalies of the input data. This method outperforms other comparison models by conducting experiments on four publicly available datasets.

Graph relationship-driven label coded mapping and compensation for multi-label textile fiber recognition

Recently, neural network-based methods for multi-label textile fiber recognition have achieved considerable success. However, a significant limitation of most current approaches is their disregard for the valuable dependencies that exist among different fiber categories. The universal multi-label image recognition methods that consider label relationships often fall short when applied to the challenge posed by the mixture of fibers in textile images. And these relationship modeling manners have not yet been used in the field of multi-label textile fiber recognition. In this work, a graph relationship-driven method is proposed for the recognition of multi-label textile fibers. Based on the graph attention network, the proposed method introduces global relation graph, label coded mapping, and label compensation to equip the feature learning backbone with the relationship modeling ability. First, feature learning backbone extracts the semantic features from images, and global relation graph mines shallow representations of global label relationships from label embeddings. Then, label coded mapping combines these semantic features and global features to model joint relationships. The obtained joint representations constitute the label code, which is used to map the fiber class indices. Finally, label compensation further extracts the deep representations of global label relationships and utilizes them to compensate the fiber class indices to obtain final prediction scores. By employing experiments on a textile fiber dataset and a rearranged dataset, the effectiveness and superiority of the proposed method are validated. Further experiments on PASCAL VOC 2007 showcase the potential applications of the proposed method beyond textile fiber recognition.

Floorplanning with Edge-aware Graph Attention Network and Hindsight Experience Replay

In this article, we focus on chip floorplanning, which aims to determine the location and orientation of circuit macros simultaneously, so the chip area and wirelength are minimized. As the highest level of abstraction in hierarchical physical design, floorplanning bridges the gap between the system-level design and the physical synthesis, whose quality directly influences downstream placement and routing. To tackle chip floorplanning, we propose an end-to-end reinforcement learning (RL) methodology with a hindsight experience replay technique. An edge-aware graph attention network (EAGAT) is developed to effectively encode the macro and connection features of the netlist graph. Moreover, we build a hierarchical decoder architecture mainly consisting of transformer and attention pointer mechanism to output floorplan actions. Since the RL agent automatically extracts knowledge about the solution space, the previously learned policy can be quickly transferred to optimize new unseen netlists. Experimental results demonstrate that, compared with state-of-the-art floorplanners, the proposed end-to-end methodology significantly optimizes area and wirelength on public GSRC and MCNC benchmarks.

Abstract PO4-07-04: Breast Cancer HER2 Status Prediction from Hematoxylin-Eosin Stained Images Using Point Cloud Transformer

Abstract Background: Human epidermal growth factor receptor-2 (HER2) status is a key factor in determining the treatment strategy for breast cancer patients. Patients with HER2-positive status are more likely to benefit from HER2-targeted therapy, leading to improved prognosis. In current routine diagnostic practice, pathologists utilize Hematoxylin-Eosin (HE) stained tumor tissues for histopathological assessment. Subsequently, IHC assessment and/or FISH test are performed to evaluate the HER2 status. However, manual assessment results may be affected by tissue usability and observer-subjectivity. Therefore, there is a necessity to predict the HER2 status directly from HE images to minimize time and cost while ensuring enhanced consistency. Methods: We identified 608 HE diagnostic slides with HER2 status from The Cancer Genome Atlas in breast cancer (TCGA-BRCA). It contains 474 HER2-positive slides (IHC 3+, IHC 2+ and FISH positive) and 134 HER2-negative slides (IHC 0, IHC 1+, IHC 2+ and FISH negative). To analyze these slides, we first tiled the HE images into patches with a fixed size of 256 × 256 at 20 × magnification. Then the patch-level feature was derived from a self-supervised pretrained transformer. Meanwhile, artificial intelligence (AI) methods are adopted to predict HER2 status from HE images. To capture the long-distance patch relations within a slide, we represented the patches as distinct points and utilized the Point Cloud Transformer model for HER2 status prediction. Specifically, 1024 patches (points) in each slide were randomly selected and input into the Point Cloud Transformer. This process yielded a slide-level prediction result for the HER2 status. Furthermore, Graph Attention Network (GAT), Graph Sample and Aggregate Network (Graph SAGE), and hierarchical Point Cloud Network (PointNet++) were adopted to compare the effectiveness of HER2 status prediction using Point Cloud Transformer. Of note that Point Cloud Transformer incorporated attention mechanisms for point aggregation compared to PointNet++. Results: The Point Cloud Transformer was trained with 5-fold cross-validation, and the area under the curve (AUC) of the receiver operating characteristic (ROC) curve was reported. The performance of point-based models outperformed the graph-based models. And the Point Cloud Transformer achieved the highest AUC of 0.7496 among all AI models. The detailed AUC for each AI model was shown in Table 1. Conclusion: Our study revealed that the HER2 status can be predicted directly from HE images without using IHC images. Furthermore, point-based models have demonstrated the ability to capture long-distance relations among patches, surpassing graph-based models in terms of prediction performance. To further enhance performance, we adopted a better point aggregation method, such as the Point Cloud Transformer, which held promise for further improving the accuracy of predictions in the future. Table 1: AUC of HER2 status prediction from HE images in the TCGA-BRCA dataset Method GAT Graph SAGE PointNet++ Point Cloud Transformer AUC 0.6640 0.6737 0.7201 0.7496 Citation Format: Bao Li, Zhenyu Liu, Yang Du, Jie Tian. Breast Cancer HER2 Status Prediction from Hematoxylin-Eosin Stained Images Using Point Cloud Transformer [abstract]. In: Proceedings of the 2023 San Antonio Breast Cancer Symposium; 2023 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2024;84(9 Suppl):Abstract nr PO4-07-04.

A combination network based on graph modules for cardiac segmentation

BackgroundCardiac segmentation plays a crucial role in extracting cardiac substructures from computed tomography images, providing necessary information for subsequent cardiac disease diagnostics. Current research on cardiac segmentation algorithms faces challenges such as complex spatial distributions and limited labeled samples. The preferred method, convolutional neural network (CNN), excels at capturing features at various levels. However, the coarse convolutional operations of CNN often neglect valuable structural information. In contrast, graph neural network (GNN) excels at capturing complex spatial relationships among nodes by aggregating messages from graphs. MethodologyIn this work, a GNN-CNN combination network (GCCN) with enhanced feature extraction by GNN is proposed for cardiac segmentation. The combination network integrates the strengths of two approaches while compensating for each other's shortcomings. Firstly, 3D CT images are transformed into graphs, whose nodes with similar feature values within a certain spatial proximity are connected. Secondly, a lightweight graph attention network (LGAT) module is proposed for aggregating the messages of constructed graphs. Thirdly, the data, resampled from the feature-enhanced graphs, is fed into the 3D ResU-Net for the final segmentation process. ResultThe proposed method achieves an average Dice score of 92.98% which is 1.84% higher than 3D ResU-Net. The average Dice score of each cardiac substructure improved, particularly for the ascending aorta, achieving 96.35%, an 8.29% improvement over 3D ResU-Net. Additionally, our study demonstrates that GCCN has the improved adaptability in handling datasets with limited samples. ConclusionIn general, GCCN allows for the extraction of richer structural information from CT images, resulting in stronger feature representation of the data. Compared to pure convolutional methods, GCCN not only enhances segmentation accuracy but also exhibits greater adaptability to small-sample data. GCCN represents an exploration of applying graph neural networks to cardiac segmentation tasks from a novel perspective.

Dual-channel relative position guided attention networks for aspect-based sentiment analysis

Aspect-based sentiment analysis (ABSA) aims to match sentiment tendencies for different aspects of a sentence to understand the product experience of the user. It is a pressing challenge for existing ABSA methods to synthesize sentences’ semantic relevance and syntactic dependency for more comprehensive sentiment representations. In this paper, we propose a Dual-Channel Relative Position Guided Attention Network (Dual-RPGA). Dual-RPGA deeply learns semantic and syntactic representations of sentiment to provide reliable knowledge for dynamic fusion and prediction of sentiment. First, we design a syntactic graph attention network (Syn-GAT) to learn the syntactic relative position between aspect and context, which guides the sentiment syntactic representation. Then, we build a semantic attention network (Sem-Attention). It computes semantic attention and similarity coefficients for aspects and contexts to enhance sentiment semantic expressions. Finally, we design a fusion network (Bi-Fusion) that realizes dynamic feature interactions of sentiment semantics and syntactics to perform sentiment prediction. We conduct extensive experiments on two groups of datasets to validate the performance of Dual-RPGA on the ABSA task. The results show that Dual-RPGA outperforms the optimal baseline by 0.58%∼1.49% of the Acc score, which verifies that Dual-RPGA performs better on the ABSA task.

Comprehensive database for food-gut microbiota-disease interactions (FGMDI) analysis and dietary recommendation applications

Regulating gut microbiota through dietary recommendations is crucial in health management. Traditional approaches for identifying diet–microbe–disease links are often resource-intensive and cumbersome, thus hampering timely and personalized regulatory efforts. By combining existing databases and related literature, this paper proposes Food–Gut Microbiota–Disease Interactions (FGMDI), an extensive database that encompasses 313 foods, 1532 gut microbes, and 281 disease entities, thus resulting in 1806 food–gut microbe associations and 11,063 disease–gut microbe associations. Moreover, we developed a predictive model powered by graph attention networks to predict food–microbe and disease–microbe associations. Finally, we go beyond prediction by offering tailored dietary recommendations. By seamlessly merging quantitative model insights with qualitative human metagenomic experimental data, we honed our predictions to enhance precision and relevance. Our model achieved remarkable area under the curve (AUC) of 0.985 and 0.986 in cross-validation, thus demonstrating its excellent accuracy and efficacy. And these personalized dietary recommendations address various conditions, including type 2 diabetes mellitus, ulcerative colitis, and obesity, and are closely aligned with the existing literature. FGMDI is a comprehensive resource that covers the landscape of microbiota analysis and dietary recommendations, an indispensable tool for personal health management.

Aspect-Aware Graph Attention Network for Heterogeneous Information Networks.

Graph Convolutional Networks (GCNs) derive inspiration from recent advances in computer vision, by stacking layers of first-order filters followed by a nonlinear activation function to learn entity or graph embeddings. Although GCNs have been shown to boost the performance of many network analysis tasks, they still face tremendous challenges in learning from Heterogeneous Information Networks (HINs), where relations play a decisive role in knowledge reasoning. What's more, there are multiaspect representations of entities in HINs, and a filter learned in one aspect do not necessarily apply to another. We address these challenges by proposing the Aspect-Aware Graph Attention Network (AGAT), a model that extends GCNs with alternative learnable filters to incorporate entity and relational information. Instead of focusing on learning the general entity embeddings, AGAT learns the adaptive entity embeddings based on prediction scenario. Experiments of link prediction and semi-supervised classification verify the effectiveness of our algorithm.

Co-Guiding for Multi-Intent Spoken Language Understanding.

Recent graph-based models for multi-intent SLU have obtained promising results through modeling the guidance from the prediction of intents to the decoding of slot filling. However, existing methods (1) only model the unidirectional guidance from intent to slot, while there are bidirectional inter-correlations between intent and slot; (2) adopt homogeneous graphs to model the interactions between the slot semantics nodes and intent label nodes, which limit the performance. In this paper, we propose a novel model termed Co-guiding Net, which implements a two-stage framework achieving the mutual guidances between the two tasks. In the first stage, the initial estimated labels of both tasks are produced, and then they are leveraged in the second stage to model the mutual guidances. Specifically, we propose two heterogeneous graph attention networks working on the proposed two heterogeneous semantics-label graphs, which effectively represent the relations among the semantics nodes and label nodes. Besides, we further propose Co-guiding-SCL Net, which exploits the single-task and dual-task semantics contrastive relations. For the first stage, we propose single-task supervised contrastive learning, and for the second stage, we propose co-guiding supervised contrastive learning, which considers the two tasks' mutual guidances in the contrastive learning procedure. Experiment results on multi-intent SLU show that our model outperforms existing models by a large margin, obtaining a relative improvement of 21.3% over the previous best model on MixATIS dataset in overall accuracy. We also evaluate our model on the zero-shot cross-lingual scenario and the results show that our model can relatively improve the state-of-the-art model by 33.5% on average in terms of overall accuracy for the total 9 languages.

Promoting knowledge recommendation in innovative engineering design: a BERT-GAT-based patent representation learning approach

Since innovation in complex product design hinges on thorough engineering knowledge application, high-quality patent recommendations foster innovation in engineering design. However, many patent knowledge recommendation studies perform patent analysis without comprehensive exploration and proper organisation of knowledge, causing a superficial understanding of patents and returning arbitrary results. To mitigate this issue, a deep learning-based approach for patent representation learning and knowledge recommendation is proposed. First, a four-dimensional patent knowledge model is defined to formalise the patent attributes that critically affect the engineering design outcomes, namely patents’ domain(D), function(F), technology(T) and citation(C). Second, to exploit patent knowledge from their content and citation relationships, a representation learning approach integrating Bidirectional Encoder Representations from Transformers(BERT) and Graph Attention Network(GAT) is introduced. Thereafter a patent knowledge space is established in which each patent is characterised by the function, technology, and citation embeddings. Third, a knowledge requirement space is also constructed by vectorising a designer’s search query via BERT model and linking it to a requirement-representing patent based on similarity. Finally, a recommender prototype is developed and showcased by the knowledge recommendation in sealing structure design tasks. Comparative experiments and application cases validate the effectiveness of our method in patent representation learning and knowledge recommendation.

A Study of Adjacent Intersection Correlation Based on Temporal Graph Attention Network.

Traffic state classification and relevance calculation at intersections are both difficult problems in traffic control. In this paper, we propose an intersection relevance model based on a temporal graph attention network, which can solve the above two problems at the same time. First, the intersection features and interaction time of the intersections are regarded as input quantities together with the initial labels of the traffic data. Then, they are inputted into the temporal graph attention (TGAT) model to obtain the classification accuracy of the target intersections in four states-free, stable, slow moving, and congested-and the obtained neighbouring intersection weights are used as the correlation between the intersections. Finally, it is validated by VISSIM simulation experiments. In terms of classification accuracy, the TGAT model has a higher classification accuracy than the three traditional classification models and can cope well with the uneven distribution of the number of samples. The information gain algorithm from the information entropy theory was used to derive the average delay as the most influential factor on intersection status. The correlation from the TGAT model positively correlates with traffic flow, making it interpretable. Using this correlation to control the division of subareas improves the road network's operational efficiency more than the traditional correlation model does. This demonstrates the effectiveness of the TGAT model's correlation.

A neuroergonomics model for evaluating nuclear power plants operators’ performance under heat stress driven by ECG time-frequency spectrums and fNIRS prefrontal cortex network: A CNN-GAT fusion model

Operators experience complicated physiological and psychological states when exposed to extreme heat stress, which can impair cognitive function and decrease performance significantly, ultimately leading to severe secondary disasters. Therefore, there is an urgent need for a feasible technique to identify their abnormal states to enhance the reliability of human-cybernetics systems. With the advancement of deep learning in physiological modeling, a model for evaluating operators’ performance driven by electrocardiogram (ECG) and functional near-infrared spectroscopy (fNIRS) was proposed, demonstrating high ecological validity. The model fused a convolutional neural network (CNN) backbone and a graph attention network (GAT) backbone to extract discriminative features from ECG time-frequency spectrums and fNIRS prefrontal cortex (PFC) network respectively with deeper neuroscience domain knowledge, and eventually achieved 0.90 AUC. Results supported that handcrafted features extracted by specialized neuroscience methods can alleviate overfitting. Inspired by the small-world nature of the brain network, the fNIRS PFC network was organized as an undirected graph and embedded by GAT. It is proven to perform better in information aggregation and delivery compared to a simple non-linear transformation. The model provides a potential neuroergonomics application to evaluate the human state in vital human-cybernetics systems under industry 5.0 scenarios.

Modification and completion of geological structure knowledge graph based on pattern matching

As a knowledge representation method, knowledge graph is widely used in intelligent question answering systems and recommendation systems. At present, the research on knowledge graph mainly focuses on information query and retrieval based on knowledge graph. In some domain knowledge graphs, specific subgraph structures (patterns) have specific physical meanings. Aiming at this problem, this paper proposes a method and framework of knowledge graph pattern mining based on gat. Firstly, the patterns with specific physical meaning were transformed into subgraph structures containing topological structures and entity attributes. Secondly, the subgraph structure of the pattern is regarded as the query graph, and the knowledge graph is regarded as the data graph, so that the problem is transformed into an approximate subgraph matching problem. Then, the improved relational graph attention network is used to fuse the adaptive edge deletion mechanism to realize the approximate subgraph matching of subgraph structure and attribute, so as to obtain the best matching subgraph. The proposed method is trained in an end-to-end manner. The approximate subgraph matching is realized on the existing data set, and the research work of key pattern mining of complex geological structure knowledge graph is carried out.

A dynamic graph deep learning model with multivariate empirical mode decomposition for network‐wide metro passenger flow prediction

AbstractNetwork‐wide short‐term passenger flow prediction is critical for the operation and management of metro systems. However, it is challenging due to the inherent non‐stationarity, nonlinearity, and spatial–temporal dependencies within passenger flow. To tackle these challenges, this paper introduces a hybrid model called multi‐scale dynamic propagation spatial–temporal network (MSDPSTN). Specifically, the model employs multivariate empirical mode decomposition to jointly decompose the multivariate passenger flow into multi‐scale intrinsic mode functions. Then, a set of dynamic graphs is developed to reveal the passenger propagation law in metro networks. Based on the representation, a deep learning model is proposed to achieve multistep passenger flow prediction, which employs the dynamic propagation graph attention network with long short‐term memory to extract the spatial–temporal dependencies. Extensive experiments conducted on a real‐world dataset from Chengdu, China, validate the superiority of the proposed model. Compared to state‐of‐the‐art baselines, MSDPSTN reduces the mean absolute error, root mean squared error, and mean absolute percentage error by at least 3.243%, 4.451%, and 4.139%, respectively. Further quantitative analyses confirm the effectiveness of the components in MSDPSTN. This paper contributes to addressing inherent features of passenger flow to enhance prediction performance, offering critical insights for decision‐makers in implementing real‐time operational strategies.

Spatiotemporal synchronous dynamic graph attention network for traffic flow forecasting

Spatiotemporal synchronous dynamic graph attention network for traffic flow forecasting

Semi-supervised few-shot fault diagnosis driven by multi-head dynamic graph attention network under speed fluctuations

Although graph neural networks (GNNs) have achieved great success in the field of semi-supervised bearing fault diagnosis recently, most GNN-based studies still face the challenge of insufficient labeled samples. Meanwhile, the signals collected from bearings usually have a time offset due to the alternating speed in the actual working conditions, reducing the applicability of traditional GNN diagnosis methods relying on Euclidean distance. And incomplete use of unlabeled data and insufficient exploitation of neighborhood information further reduce the reliability of fault diagnosis models. To solve these problems, a semi-supervised few-shot fault diagnosis method driven by multi-head dynamic graph attention network (MHDGAT) is proposed in this study. First, dynamic time warping (DTW) similarity is introduced in the construction of K-nearest neighbor graph to solve the problem of local time offset caused by speed change. Second, some pseudo labels generated by smoothness assumption are injected into the raw training set, and more comprehensive information can be learned in the process of model training. Then, the presented MHDGAT model can mine rich fault characteristics from different scales by the multi-head attention and extract the most sensitive features due to the proposed dynamic graph mechanism adaptively. Finally, the fault categories can be obtained by a softmax classifier located at the end of the model. The proposed model has been validated in two case studies, and the experimental results indicate that MHDGAT exerts strong performance even under extremely low-label ratios.

Predict lncRNA-drug associations based on graph neural network.

LncRNAs are an essential type of non-coding RNAs, which have been reported to be involved in various human pathological conditions. Increasing evidence suggests that drugs can regulate lncRNAs expression, which makes it possible to develop lncRNAs as therapeutic targets. Thus, developing in-silico methods to predict lncRNA-drug associations (LDAs) is a critical step for developing lncRNA-based therapies. In this study, we predict LDAs by using graph convolutional networks (GCN) and graph attention networks (GAT) based on lncRNA and drug similarity networks. Results show that our proposed method achieves good performance (average AUCs > 0.92) on five datasets. In addition, case studies and KEGG functional enrichment analysis further prove that the model can effectively identify novel LDAs. On the whole, this study provides a deep learning-based framework for predicting novel LDAs, which will accelerate the lncRNA-targeted drug development process.

Predicting stock price crash risk in China: A modified graph WaveNet model

The stock price of a firm is dynamically influenced by its own factors as well as those of its peers. In this study, we introduce a Graph Attention Network (GAT) integrated with WaveNet architecture—termed the GAT-WaveNet model—to capture both time-series and spatial dependencies for forecasting the stock price crash risk of Chinese listed firms from 2012 to 2023. Utilizing node-rolling techniques to prevent overfitting, our results show that the GAT-WaveNet model significantly outperforms traditional machine learning models in prediction accuracy. Moreover, investment portfolios leveraging the GAT-WaveNet model substantially exceed the cumulative returns of those based on other models.

EA-GAT: Event aware graph attention network on cyber-physical systems

Anomaly detection with high accuracy, recall, and low error rate is critical for the safe and uninterrupted operation of cyber-physical systems. However, detecting anomalies in multimodal time series with different modalities obtained from cyber-physical systems is challenging. Although deep learning methods show very good results in anomaly detection, they fail to detect anomalies according to the requirements of cyber-physical systems. In the use of graph-based methods, data loss occurs during the conversion of time series into graphs. The fixed window size used to transform time series into graphs causes a loss of spatio-temporal correlations. In this study, we propose an Event Aware Graph Attention Network (EA-GAT), which can detect anomalies by event-based cyber-physical system analysis. EA-GAT detects and tracks the sensors in cyber-physical systems and the correlations between them. The system analyzes and models the relationship between the components during the marked periods as a graph. Anomalies in the system are found through the created graph models. Experiments show that the EA-GAT technique is more effective than other deep learning methods on SWaT, WADI, MSL datasets used in various studies. The event-based dynamic approach is significantly superior to the fixed-size sliding window technique, which uses the same learning structure. In addition, anomaly analysis is used to identify the attack target and the affected components. At the same time, with the slip subsequence module, the data is divided into subgroups and processed simultaneously.