Search

With the increasing need for public health and drug development, combination therapy has become widely used in clinical settings. However, the risk of unanticipated adverse effects and unknown toxicity caused by drug-drug interactions (DDIs) is a serious public health issue for polypharmacy safety. Traditional experimental methods for detecting DDIs are expensive and time-consuming. Therefore, many computational methods have been developed in recent years to predict DDIs with the growing availability of data and advancements in artificial intelligence. In silico methods have proven to be effective in predicting DDIs, but detecting potential interactions, especially for newly discovered drugs without an existing DDI network, remains a challenge. In this study, we propose a predicting method of DDIs named HAG-DDI based on graph attention networks. We consider the differences in mechanisms between DDIs and add learning of semantic-level attention, which can focus on advanced representations of DDIs. By treating interactions as nodes and the presence of the same drug as edges, and constructing small subnetworks during training, we effectively mitigate potential bias issues arising from limited data availability. Our experimental results show that our method achieves an F1-score of 0.952, proving that our model is a viable alternative for DDIs prediction. The codes are available at: https://github.com/xtnenu/DDIFramework.

Abstract The expansion of available information in location-based social networks (LBSNs) has led to information overload, making it urgent to discover users’ next point-of-interest (POI). Some existing works only consider certain modal information in LBSNs and do not transform them into high-dimensional structures, which hinders the alleviation of the data sparsity problem. Moreover, many approaches rely solely on social relationships, making it difficult to recommend POIs to new users without association information. To tackle these challenges, we propose a social- and spatial–temporal-aware next point-of-Interest (SSTP) recommendation model. SSTP uses two feature encoders based on self-attention mechanism and gate recurrent unit to model users’ check-in enhancement sequence hierarchically. We also design a random neighborhood sampling approach to mine user social relationships, thus alleviating the user cold start problem. Finally, we propose a geographical-aware graph attention network to learn the sensitivity of users to distance. Extensive experiments on two real-world datasets show that SSTP outperforms state-of-the-art models, improving Hit@k by 2.26–6.55 $$\%$$ % and MAP@k by 3.49–6.55 $$\%$$ % . Moreover, SSTP has better performance on sparse data, with an average improvement of 6.09 $$\%$$ % on the Hit@k. The code can be downloaded at https://github.com/Rih0/sstp .

Abstract With the widespread use of blockchain, more and more smart contracts are being deployed, and their internal logic is getting more and more sophisticated. Due to the large false positive rate and low detection accuracy of most current detection methods, which heavily rely on already established detection criteria, certain smart contracts additionally call for human secondary detection, resulting in low detection efficiency. In this study, we propose HGAT, a hierarchical graph attention network-based detection model, in order to address the aforementioned issues as well as the shortcomings of current smart contract vulnerability detection approaches. First, using Abstract Syntax Tree (AST) and Control Flow Graph, the functions in the smart contract are abstracted into code graphs (CFG). Then abstract each node in the code subgraph, extract the node features, utilize the graph attention mechanism GAT, splice the obtained vectors to form the features of each line of statements and use these features to detect smart contracts. To create test data and assess HGAT, we leverage the open-source smart contract vulnerability sample dataset. The findings of the experiment indicate that this method can identify smart contract vulnerabilities more quickly and precisely than other detection techniques.

Abstract Currently, the Internet is widely used for shopping. Online reviews have become a crucial factor in helping people to make purchasing decisions. However, the large amount of data overwhelms most users, leading to the problem of information overload. To address this issue, researchers have proposed recommender systems as a solution. The most commonly used method is the collaborative filtering method, which analyzes users’ purchase history or behavior to make recommendations. In addition to user ratings, by analyzing users’ comments and the relationships between users and items more precise preferences can be obtained. In this study, the aspect-based rating prediction with a hybrid deep learning method (ARPH) is proposed. It consists of five parts: aspect detection, sentiment and semantic analysis, user preference analysis, graph attention network analysis, and rating prediction. It initially extracts the implicit aspect features and aspects’ sentiment–semantic features from user and item reviews. The convolutional neural network and matrix factorization methods are then used to generate the predicted ratings of items. Additionally, a graph attention network was built to calculate the predicted ratings based on the relationships between users and items. Finally, a multilayer perceptron was used to automatically adjust the weights for integrating these two predicted ratings. Our method utilizes user–item relationships to predict ratings when there are fewer user reviews. Conversely, the features derived from textual reviews were employed for rating prediction. The experimental results showed that extracting different features is useful in representing user and product preferences. The proposed method effectively improved the accuracy of the rating predictions.

Accurately estimating commuting flow is essential for optimizing urban planning and traffic design. The latest graph neural network (GNN) model with the encoder-decoder-predictor components has several limitations. First, it ignores the temporal dependency of node features for node embedding. Second, different estimation methods used in the decoder and predictor make it difficult to distinguish the contribution of node embedding or estimation method to flow estimation. Third, finer-grained socio-economic features of nodes are difficult to obtain due to low data availability. To address these problems, this study proposes a fusion model of temporal graph attention network and machine learning (TGAT-ML) to infer commuting flow from dynamic human activity intensity distribution. The model first constructs a commuting network with temporal human activity intensity as node features. A temporal graph attention network is then developed to capture the spatiotemporal dependency. The learned node embedding is generated by using a machine learning method in the decoder. Finally, based on learned node embedding and machine learning method used in the decoder, the commuting flow intensity is estimated. Results from an empirical study using the Baidu heat map data of Guangzhou city indicate that the proposed fusion model TGAT-ML outperforms all other baseline models. This study proves that the model performance can be significantly enhanced by determining the edge existence through commuting time-based approach, integrating temporal convolution with graph convolution, and unifying flow estimation method in both decoder and predictor. This work enables commuting flow estimation from dynamic human activity intensity and broadens existing flow generation research in terms of data and methodology.

In this paper, we utilize big data to screen relevant data on charging safety influencing factors and perform data cleaning to constitute a charging safety influencing factors dataset. BERT is selected as the baseline model for the named entity recognition task, together with the CRF model, to exclude irrelevant features, resulting in an effective model for entity recognition in line with the knowledge graph. Introducing a security database, a graph attention network model that simultaneously obtains the structural features and textual description features of the security knowledge graph is proposed to improve the performance of knowledge graph relationship extraction. The dataset of high-frequency charging security composition, as well as the random dataset, are used as experimental samples, respectively, to compare and analyze the performance of the BERT-CRF named entity recognition model in terms of each index. The link prediction evaluation task is evaluated using the structure- and text-based graph attention network model, and experimental analysis is carried out using three benchmark models. From the overall results of the test, it can be seen that the BERT-CRF model learns 90% of the lexicon’s knowledge and passes the model test by keeping each evaluation metric in the range of 0.9 to 1.0 under the large data volume experimental environment. The proposed graph attention network model, which uses structure and text, has a better link prediction performance than other models and performs better in the FB15K-237 dataset.

Physical education teaching mode is the link between physical education teaching theory and physical education teaching practice, which has a direct impact on the quality and teaching effect of physical education teaching. In this paper, the recognition algorithm is applied to physical education, and a graph-convolutional neural network is constructed. In order to solve the problem that a graph convolutional neural network can only deal with non-motion graphs, using the Masked Attention mechanism, assigning different weights to neighboring nodes, extracting node features and features between neighboring nodes, thus the graph attention neural network, and introducing the YOLO model to predict the single and multiple postures and correct the students’ technical movements. Through the comparative experimental study in this paper, it was found that the p-values of the four dimensions of students’ attention to athletics, adverse interest in athletics learning, positive interest in athletics learning, and daily athletics exercise were all 0.000

In the context of the significant data era, this paper explores the educational applications based on diagnostic learning analytics technology to improve personalized learning and teaching effects in the educational process. The study adopts a multidimensional feature fusion approach to construct a cognitive diagnostic model to predict learners’ knowledge status and future learning performance. Through actual data testing, the model can effectively predict the students’ knowledge mastery state and analyze the students’ learning process in depth. The experimental results show that the diagnostic model exhibits high efficiency and accuracy in predicting students’ knowledge mastery status, with an accuracy rate of 92.97%, significantly better than traditional teaching methods. In addition, the study explores the encoding method of learners’ multidimensional features and constructs a dynamic diagnostic model of test factors and student factors based on graph attention network. The study provides a new learning analysis and diagnostic method in the education field, which helps improve the effect of personalized learning.

At present, there are problems such as low fault data, insufficient labeling information, and poor fault diagnosis in the field of ship engine diagnosis. To address the above problems, this paper proposes a fault diagnosis method based on probabilistic similarity and rank-order similarity of multi-head graph attention neural networks (MPGANN) models. Firstly, the ship engine dataset is used to explore the similarity between the data using the probabilistic similarity of T_SNE and the rank order similarity of Spearman's correlation coefficient to define the neighbor relationship between the samples, and then the appropriate weights are selected for the early fusion of the two graph structures to fuse the feature information of the two scales. Finally, the graph attention neural networks (GANN) incorporating the multi-head attention mechanism are utilized to complete the fault diagnosis. In this paper, comparative experiments such as graph construction and algorithm performance are carried out based on the simulated ship engine dataset, and the experimental results show that the MPGANN outperforms the comparative methods in terms of accuracy, F1 score, and total elapsed time, with an accuracy rate of 97.58%. The experimental results show that the model proposed in this paper can still fulfill the ship engine fault diagnosis task well under unfavorable conditions such as small samples and insufficient label information, which is of practical significance in the field of intelligent ship cabins and fault diagnosis. [ABSTRACT FROM AUTHOR]

In recent years, with the advancement of natural language processing techniques and the release of models like ChatGPT, how language models understand questions has become a hot topic. In handling complex logical reasoning with pre-trained models, its performance still has room for improvement. Inspired by DAGN, we propose an improved DaGATN (Discourse-apperceptive Graph Attention Networks) model. By constructing a discourse information graph to learn logical clues in the text, we decompose the context, question, and answer into elementary discourse units (EDUs) and connect them with discourse relations to construct a relation graph. The text features are learned through a discourse graph attention network and applied to downstream multiple-choice tasks. Our method was evaluated on the ReClor dataset and achieved an accuracy of 74.3%, surpassing the best-known performance methods utilizing deberta-xlarge-level pre-trained models, and also performed better than ChatGPT (Zero-Shot). [ABSTRACT FROM AUTHOR]

Drug–drug interaction (DDI) prediction can discover potential risks of drug combinations in advance by detecting drug pairs that are likely to interact with each other, sparking an increasing demand for computational methods of DDI prediction. However, existing computational DDI methods mostly rely on the single-view paradigm, failing to handle the complex features and intricate patterns of DDIs due to the limited expressiveness of the single view. To this end, we propose a H ierarchical T riple-view C ontrastive L earning framework for D rug– D rug I nteraction prediction (HTCL-DDI), leveraging the molecular, structural and semantic views to model the complicated information involved in DDI prediction. To aggregate the intra-molecular compositional and structural information, we present a dual attention-aware network in the molecular view. Based on the molecular view, to further capture inter-molecular information, we utilize the one-hop neighboring information and high-order semantic relations in the structural view and semantic view , respectively. Then, we introduce contrastive learning to enhance drug representation learning from multifaceted aspects and improve the robustness of HTCL-DDI. Finally, we conduct extensive experiments on three real-world datasets. All the experimental results show the significant improvement of HTCL-DDI over the state-of-the-art methods, which also demonstrates that HTCL-DDI opens new avenues for ensuring medication safety and identifying synergistic drug combinations. [ABSTRACT FROM AUTHOR]

Extracting social relationships between people from images has an important role in criminal investigation, privacy protection and other fields. Existing graph modeling approaches have achieved good results by creating interpersonal relationship graphs or constructing knowledge graphs to learn people's relationships. However, the methods based on graph convolutional neural network (GCN) ignore different degrees of importance of different features for specific relationships to some extent. In order to solve this problem, this paper proposed a graph attention-based double-branch social relationship recognition model(GAT-DBSR). The first branch extracting person regions as well as image global features as nodes, and the core updated these nodes to learn feature representations of person relationships through graph attention networks and gating mechanisms. The second branch extracted scene features by convolutional neural networks to enhance the recognition of relationships between people. Finally, it fused and classified the features of the two branches to obtain all social relationships. The model achieves an mAP of 74.4% on the fine-grained relationship recognition task on the PISC dataset, a 1.2% improvement compared to the baseline model. The accuracy of relationship recognition on the PIPA dataset also shows some improvement. The experimental results show that the model has better results. [ABSTRACT FROM AUTHOR]

Copyright of Journal of Jilin University (Science Edition) / Jilin Daxue Xuebao (Lixue Ban) is the property of Zhongguo Xue shu qi Kan (Guang Pan Ban) Dian zi Za zhi She and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Monocular Visual Odometry (VO) is often formulated as a sequential dynamics problem that relies on scene rigidity assumption. One of the main challenges is rejecting moving objects and estimating camera pose in dynamic environments. Existing methods either take the visual cues in the whole image equally or eliminate the fixed semantic categories by heuristics or attention mechanisms. However, they fail to tackle unknown dynamic objects which are not labeled in the training sets of the network. To solve these issues, we propose a novel framework, named graph attention network (GAT)-optimized dynamic monocular visual odometry (GDM-VO), to remove dynamic objects explicitly with semantic segmentation and multi-view geometry in this paper. Firstly, we employ a multi-task learning network to perform semantic segmentation and depth estimation. Then, we reject priori known and unknown objective moving objects through semantic information and multi-view geometry, respectively. Furthermore, to our best knowledge, we are the first to leverage GAT to capture long-range temporal dependencies from consecutive image sequences adaptively, while existing sequential modeling approaches need to select information manually. Extensive experiments on the KITTI and TUM datasets demonstrate the superior performance of GDM-VO overs existing state-of-the-art classical and learning-based monocular VO. [ABSTRACT FROM AUTHOR]

A recommendation algorithm combined with a knowledge graph enables auxiliary information on items to be obtained by using the knowledge graph to achieve better recommendations. However, the recommendation performance of existing methods relies heavily on the quality of the knowledge graph. Knowledge graphs often contain noise and irrelevant connections between items and entities in the real world. This knowledge graph sparsity and noise significantly amplifies the noise effects and hinders the accurate representation of user preferences. In response to these problems, an improved collaborative recommendation model is proposed which integrates knowledge embedding and graph contrastive learning. Specifically, we propose a knowledge contrastive learning scheme to mitigate noise within the knowledge graph during information aggregation, thereby enhancing the embedding quality of items. Simultaneously, to tackle the issue of insufficient user-side information in the knowledge graph, graph convolutional neural networks are utilized to propagate knowledge graph information from the item side to the user side, thereby enhancing the personalization capability of the recommendation system. Additionally, to resolve the over-smoothing issue in graph convolutional networks, a residual structure is employed to establish the message propagation network between adjacent layers of the same node, which expands the information propagation path. Experimental results on the Amazon-book and Yelp2018 public datasets demonstrate that the proposed model outperforms the best baseline models by 11.4% and 11.6%, respectively, in terms of the Recall@20 evaluation metric. This highlights the method's efficacy in improving the recommendation accuracy and effectiveness when incorporating knowledge graphs into the recommendation process. [ABSTRACT FROM AUTHOR]

The carcinogenicity of drugs can have a serious impact on human health, so carcinogenicity testing of new compounds is very necessary before being put on the market. Currently, many methods have been used to predict the carcinogenicity of compounds. However, most methods have limited predictive power and there is still much room for improvement. In this study, we construct a deep learning model based on capsule network and attention mechanism named DCAMCP to discriminate between carcinogenic and non‐carcinogenic compounds. We train the DCAMCP on a dataset containing 1564 different compounds through their molecular fingerprints and molecular graph features. The trained model is validated by fivefold cross‐validation and external validation. DCAMCP achieves an average accuracy (ACC) of 0.718 ± 0.009, sensitivity (SE) of 0.721 ± 0.006, specificity (SP) of 0.715 ± 0.014 and area under the receiver‐operating characteristic curve (AUC) of 0.793 ± 0.012. Meanwhile, comparable results can be achieved on an external validation dataset containing 100 compounds, with an ACC of 0.750, SE of 0.778, SP of 0.727 and AUC of 0.811, which demonstrate the reliability of DCAMCP. The results indicate that our model has made progress in cancer risk assessment and could be used as an efficient tool in drug design. [ABSTRACT FROM AUTHOR]

A growing number of studies have shown that aberrant microRNA (miRNA) expression is closely associated with the evolution and development of various complex human diseases. These key biomarkers' identification and observation are significant for gaining a deeper understanding of disease pathogenesis and therapeutic mechanisms. Consequently, pinpointing potential miRNA–disease associations (MDA) has become a prominent bioinformatics subject, encouraging several new computational methods given the advances in graph neural networks (GNN). Nevertheless, these existing methods commonly fail to exploit the network nodes' global feature information, leaving the generation of high-quality embedding representations using graph properties as a critical unsolved issue. Addressing these challenges, we introduce the DAEMDA, a computational method designed to optimize the current models' efficacy. First, we construct similarity and heterogeneous networks involving miRNAs and diseases, relying on experimentally corroborated miRNA–disease association data and analogous information. Then, a newly-fashioned parallel dual-channel feature encoder, designed to better comprehend the global information within the heterogeneous network and generate varying embedding representations, follows this. Ultimately, employing a neural network classifier, we merge the dual-channel embedding representations and undertake association predictions between miRNA and disease nodes. The experimental results of five-fold cross-validation and case studies of major diseases based on the HMDD v3.2 database show that this method can generate high-quality embedded representations and effectively improve the accuracy of MDA prediction. [ABSTRACT FROM AUTHOR]

At present, the trend of familiarization of malicious code is becoming more and more obvious, and the research on the homology of malware (the classification of malicious code family) is of great significance for maintaining network security. In order to better express the overall characteristics of malicious code and improve the effect of detection and homology analysis, this paper proposes a method for detection and homology analysis of malware based on heterogeneous graphs of assembly instructions (AIHGAT). We take the assembly instructions of malicious families as the research object and analyze the importance and correlation of assembly instructions of different malicious families. The malware detection and homology analysis are carried out in three aspects: feature extraction, feature preprocessing, and model construction. In the feature extraction of malicious code, in order to alleviate the problem that it is difficult to extract static features of malicious samples that contain countermeasures such as packing and obfuscation, we obtain binary files from dynamic memory through sandbox and then, analyze its assembly instruction set. In feature preprocessing, we divide the assembly instructions into N-tuples and construct a heterogeneous graph based on assembly instructions according to the internal correlation of the gene sequence composed of the assembly N-grams features. Finally, in terms of model construction, we analyze the homology determination effect of the traditional graph neural network and construct the Graph Attention Network based on residual connection named ResGAT to analyze the homology of malicious code. The experimental results show that the ResGAT can gather the core characteristics of malicious families and enhance the ability to recognize malicious family variants. Our model has an accuracy rate of 98.83%, which is better than traditional machine learning detection methods, and can effectively determine the homology of malicious code families. [ABSTRACT FROM AUTHOR]

The agile earth observation satellite scheduling problem (AEOSSP) is a combinatorial optimization problem with time-dependent constraints. Recently, many construction heuristics and meta-heuristics have been proposed; however, existing methods cannot balance the requirements of efficiency and timeliness. In this paper, we propose a graph attention network-based decision neural network (GDNN) to solve the AEOSSP. Specifically, we first represent the task and time-dependent attitude transition constraints by a graph. We then describe the problem as a Markov decision process and perform feature engineering. On this basis, we design a GDNN to guide the construction of the solution sequence and train it with proximal policy optimization (PPO). Experimental results show that the proposed method outperforms construction heuristics at scheduling profit by at least 45%. The proposed method can also calculate the approximate profits of the state-of-the-art method with an error of less than 7% and reduce scheduling time markedly. Finally, we demonstrate the scalability of the proposed method. [ABSTRACT FROM AUTHOR]

The article focuses on enhancing news recommendation by addressing the issue of data sparsity in user behavior data. It discusses the integration of user-news relationships and user historical clicked news sequences to create a global heterogeneous transition graph and proposes the GAINRec model to improve news recommendations based on collective user behavior patterns, demonstrating its effectiveness in experiments.

Currently, computational intelligence methods, especially artificial neural networks, are increasingly applied to many scenarios. We mainly attempt to explore the task of fine-grained sentiment classification of review data through computational intelligence methods, especially artificial neural networks, and this task is also known as aspect-based sentiment analysis (ABSA). We propose a new technique called SenticGAT which is a multi-view features fusion model enhanced by an external sentiment database. We encode the external sentiment information into the syntactic dependency tree to obtain an enhanced graph with rich sentiment representation. Then we obtain multi-view features including semantics, syntactic, and sentiment features through GAT based on the enhanced graph by external knowledge. We also design a new strategy for fusing multi-view features using the feature parallel frame and convolution method. Eventually, the sentiment polarity of a specific aspect is determined based on the completely fused multi-view features. Experimental results on four public benchmark datasets demonstrate that our method is effective and sound. And it performs superiorly to existing approaches in fusion multiple-view features. [ABSTRACT FROM AUTHOR]

Copyright of Frontiers of Information Technology & Electronic Engineering is the property of Springer Nature and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Electricity load forecasting is of great significance for the overall operation of the power system and the orderly use of electricity at a later stage. However, traditional load forecasting does not consider the change in load quantity at each time point, while the information on the time difference of the load data can reflect the dynamic evolution information of the load data, which is a very important factor for load forecasting. In addition, the research topics in recent years mainly focus on the learning of the complex relationships of load sequences in time latitude by graph neural networks. The relationships between different variables of load sequences are not explicitly captured. In this paper, we propose a model that combines a differential learning network and a multidimensional feature graph attention layer, it can model the time dependence and dynamic evolution of load sequences by learning the amount of load variation at different time points, while representing the correlation of different variable features of load sequences through the graph attention layer. Comparative experiments show that the prediction errors of the proposed model have decreased by 5–26% compared to other advanced methods in the UC Irvine Machine Learning Repository Electricity Load Chart public dataset. [ABSTRACT FROM AUTHOR]

Agile-satellite mission planning is a crucial issue in the construction of satellite constellations. The large scale of remote sensing missions and the high complexity of constraints in agile-satellite mission planning pose challenges in the search for an optimal solution. To tackle the issue, a dynamic destroy deep-reinforcement learning (D3RL) model is designed to facilitate subsequent optimization operations via adaptive destruction to the existing solutions. Specifically, we first perform a clustering and embedding operation to reconstruct tasks into a clustering graph, thereby improving data utilization. Secondly, the D3RL model is established based on graph attention networks (GATs) to enhance the search efficiency for optimal solutions. Moreover, we present two applications of the D3RL model for intensive scenes: the deep-reinforcement learning (DRL) method and the D3RL-based large-neighborhood search method (DRL-LNS). Experimental simulation results illustrate that the D3RL-based approaches outperform the competition in terms of solutions' quality and computational efficiency, particularly in more challenging large-scale scenarios. DRL-LNS outperforms ALNS with an average scheduling rate improvement of approximately 11% in Area instances. In contrast, the DRL approach performs better in World scenarios, with an average scheduling rate that is around 8% higher than that of ALNS. [ABSTRACT FROM AUTHOR]

The prediction of drug–target interaction (DTI) is crucial to drug discovery. Although the interactions between the drug and target can be accurately verified by traditional biochemical experiments, the determination of DTI through biochemical experiments is a time-consuming, laborious, and expensive process. Therefore, we propose a learning-based framework named BG-DTI for drug–target interaction prediction. Our model combines two main approaches based on biological features and heterogeneous networks to identify interactions between drugs and targets. First, we extract original features from the sequence to encode each drug and target. Later, we further consider the relationships among various biological entities by constructing drug–drug similarity networks and target–target similarity networks. Furthermore, a graph convolutional network and a graph attention network in the graph representation learning module help us learn the features representation of drugs and targets. After obtaining the features from graph representation learning modules, these features are combined into fusion descriptors for drug–target pairs. Finally, we send the fusion descriptors and labels to a random forest classifier for predicting DTI. The evaluation results show that BG-DTI achieves an average AUC of 0.938 and an average AUPR of 0.930, which is better than those of five existing state-of-the-art methods. We believe that BG-DTI can facilitate the development of drug discovery or drug repurposing. [ABSTRACT FROM AUTHOR]

Ensuring that intelligent vehicles do not cause fatal collisions remains a persistent challenge due to pedestrians' unpredictable movements and behavior. The potential for risky situations or collisions arising from even minor misunderstandings in vehicle–pedestrian interactions is a cause for great concern. Considerable research has been dedicated to the advancement of predictive models for pedestrian behavior through trajectory prediction, as well as the exploration of the intricate dynamics of vehicle–pedestrian interactions. However, it is important to note that these studies have certain limitations. In this paper, we propose a novel graph-based trajectory prediction model for vehicle–pedestrian interactions called Holistic Spatio-Temporal Graph Attention (HSTGA) to address these limitations. HSTGA first extracts vehicle–pedestrian interaction spatial features using a multi-layer perceptron (MLP) sub-network and max pooling. Then, the vehicle–pedestrian interaction features are aggregated with the spatial features of pedestrians and vehicles to be fed into the LSTM. The LSTM is modified to learn the vehicle–pedestrian interactions adaptively. Moreover, HSTGA models temporal interactions using an additional LSTM. Then, it models the spatial interactions among pedestrians and between pedestrians and vehicles using graph attention networks (GATs) to combine the hidden states of the LSTMs. We evaluate the performance of HSTGA on three different scenario datasets, including complex unsignalized roundabouts with no crosswalks and unsignalized intersections. The results show that HSTGA outperforms several state-of-the-art methods in predicting linear, curvilinear, and piece-wise linear trajectories of vehicles and pedestrians. Our approach provides a more comprehensive understanding of social interactions, enabling more accurate trajectory prediction for safe vehicle navigation. [ABSTRACT FROM AUTHOR]

s t r i n g U t i l s. c o n v e r t A b s t r a c t M a t h H t m l (formulaUtilTools.convertMathHtml( s t r i n g U t i l s. c o n v e r t M m l (!article.abstractinfoEn))) With the increasing need for public health and drug development, combination therapy has become widely used in clinical settings. However, the risk of unanticipated adverse effects and unknown toxicity caused by drug-drug interactions (DDIs) is a serious public health issue for polypharmacy safety. Traditional experimental methods for detecting DDIs are expensive and time-consuming. Therefore, many computational methods have been developed in recent years to predict DDIs with the growing availability of data and advancements in artificial intelligence. In silico methods have proven to be effective in predicting DDIs, but detecting potential interactions, especially for newly discovered drugs without an existing DDI network, remains a challenge. In this study, we propose a predicting method of DDIs named HAG-DDI based on graph attention networks. We consider the differences in mechanisms between DDIs and add learning of semantic-level attention, which can focus on advanced representations of DDIs. By treating interactions as nodes and the presence of the same drug as edges, and constructing small subnetworks during training, we effectively mitigate potential bias issues arising from limited data availability. Our experimental results show that our method achieves an F1-score of 0.952, proving that our model is a viable alternative for DDIs prediction. The codes are available at: https://github.com/xtnenu/DDIFramework. [ABSTRACT FROM AUTHOR]

Copyright of Journal of Chongqing University of Technology (Natural Science) is the property of Chongqing University of Technology and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

This paper proposes a combined approach of deep deterministic policy gradient (DDPG) and graph attention network (GAT) to the geometry optimization of latticed shells with surface shapes defined by a Bézier control net. The optimization problem is formulated to minimize the strain energy of the latticed structures with heights of the Bézier control points as design variables. The information of the latticed shells, including nodal configurations, element properties and internal forces, and the Bézier control net, consisting of control points and control net, are represented as graphs using node feature matrices, adjacency matrices, and weighted adjacency matrices. A specifically designed DDPG agent utilizes GAT and matrix manipulations to observe the state of the structure through the graphs, and decides which and how Bézier control points to move. The agent is trained to excel in the task through a reward signal computed from changes in the strain energy in each optimization step. As shown in numerical examples, the trained agent can effectively optimize structures of different sizes, control nets, configurations, and initial geometries from those used during the training. The performance of the trained agent is competitive compared to particle swarm optimization and simulated annealing despite using a lower computational cost. Highlights: - A method using a reinforcement learning agent is proposed to optimize the geometry of latticed structures. - The agent is designed to observe the structure and Bézier control net and modify the Bézier control net. - The method yields good results using fewer computations when compared to other conventional methods. [ABSTRACT FROM AUTHOR]

Integrating lexical information into Chinese character embedding is a valid method to figure out the Chinese named entity recognition (NER) issue. However, most existing methods focus only on the discovery of named entity boundaries, considering only the words matched by the Chinese characters. They ignore the association between Chinese characters and their left and right matching words. They ignore the local semantic information of the character's neighborhood, which is crucial for Chinese NER. The Chinese language incorporates a significant number of polysemous words, meaning that a single word can possess multiple meanings. Consequently, in the absence of sufficient contextual information, individuals may encounter difficulties in comprehending the intended meaning of a text, leading to the emergence of ambiguity. We consider how to handle the issue of entity ambiguity because of polysemous words in Chinese texts in different contexts more simply and effectively. We propose in this paper the use of graph attention networks to construct relatives among matching words and neighboring characters as well as matching words and adding left- and right-matching words directly using semantic information provided by the local lexicon. Moreover, this paper proposes a short-sequence convolutional neural network (SSCNN). It utilizes the generated shorter subsequence encoded with the sliding window module to enhance the perception of local information about the character. Compared with the widely used Chinese NER models, our approach achieves 1.18%, 0.29%, 0.18%, and 1.1% improvement on the four benchmark datasets Weibo, Resume, OntoNotes, and E-commerce, respectively, and proves the effectiveness of the model. [ABSTRACT FROM AUTHOR]

Currently, existing research on deploying deep reinforcement learning on software-defined networks (SDN) to achieve route optimization does not consider the network's spatial–temporal correlation globally and has yet to reach the ultimate in performance. Given the above issues, this study proposes a Proximal Policy Optimization algorithm based on the Attention mechanism and Spatio–Temporal correlation (ASTPPO) to optimize the SDN routing issue. First, we extract temporal and spatial correlation features in state information using Gated Recurrent Units (GRU) and Graph Attention Networks (GAT), providing implicit information containing more environments for reinforcement learning decisions. Second, we use the skip-connect method to connect implicit and directly related information into a multi-layer perceptron, improving the model's learning efficiency and perceptual ability. Finally, we demonstrate the effectiveness of ASTPPO through static and dynamic traffic experiments. Benefitting from Spatio–Temporal correlation learning with a global view, ASTPPO performs better load balancing and congestion control under different traffic intensity requirements and network topologies than other reinforcement learning baseline algorithms. The simulation results show that the ASTPPO algorithm improved by 9.02% and 15.07%, respectively, compared with the second-best algorithm in static and dynamic traffic scenarios. [ABSTRACT FROM AUTHOR]

Copyright of Journal on Communication / Tongxin Xuebao is the property of Journal on Communications Editorial Office and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Indices recommendation is a long-standing topic in stock market investment. Predicting the future trends of indices and ranking them based on the prediction results is the main scheme for indices recommendation. How to improve the forecasting performance is the central issue of this study. Inspired by the widely used trend-following investing strategy in financial investment, the indices' future trends are related to not only the nearby transaction data but also the long-term historical data. This article proposes the MSGraph, which tries to improve the index ranking performance by modeling the correlations of short and long-term historical embeddings with the graph attention network. The original minute-level transaction data is first synthesized into a series of K-line sequences with varying time scales. Each K-line sequence is input into a long short-term memory network (LSTM) to get the sequence embedding. Then, the embeddings for all indices with the same scale are fed into a graph convolutional network to achieve index aggregation. All the aggregated embeddings for the same index are input into a graph attention network to fuse the scale interactions. Finally, a fully connected network produces the index return ratio for the next day, and the recommended indices are obtained through ranking. In total, 60 indices in the Chinese stock market are selected as experimental data. The mean reciprocal rank, precision, accuracy and investment return ratio are used as evaluation metrics. The comparison results show that our method achieves state-of-the-art results in all evaluation metrics, and the ablation study also demonstrates that the combination of multiple scale K-lines facilitates the indices recommendation.

Knowledge tracing aims to trace students’ knowledge state(the degree of knowledge) based on their historical answer performance in real time and predict their future answer performance.The current research only explores the direct influence of the question or concept itself on the performance of students’ answering questions,while often ignores the indirect influence of the deep-level information in the questions and the concepts contained on the performance of students’ answering questions.In order to make better use of these deep-level information,a graph attention deep knowledge tracing model integrated with IRT(GAKT-IRT) is proposed,which integrates item response theory(IRT).The graph attention network is applied to the field of knowledge tracing and uses IRT to increase the interpretability of the model.First,obtain the deep-level feature representation of the problem through the graph attention network layer.Next,model students’ knowledge state based on their historical answer sequence that combines the in-depth information.Then,use IRT to predict students’ future answer performance.Results of comparative experiments on 6 open real online education datasets prove that the GAKT-IRT model can better complete the knowledge tracing task and has obvious advantages in predicting the future performance of students in answering questions.