Search Results (61)

Chemical artificial intelligence increasingly relies on molecular property prediction models trained from proprietary compound libraries, bioassay records, and reaction-screening data. However, these data often contain commercially sensitive structures, confidential activity labels, and privacy-relevant experimental metadata, making direct centralization impractical. This paper proposes PrivMol, a privacy-preserving computational chemistry framework for federated molecular representation learning. PrivMol introduces two novel algorithms: Secure Substructure-Aware Federated Optimization and Differentially Private Molecular Gradient Calibration. The first algorithm decomposes molecular graphs into privacy-sensitive and task-relevant substructure regions, enabling local clients to train graph neural networks while transmitting only securely aggregated model updates. The second algorithm adaptively calibrates clipping and perturbation according to atom- and substructure-level contribution scores, reducing unnecessary utility loss on chemically informative fragments while retaining formal differential privacy guarantees. To improve robustness under heterogeneous chemical spaces, PrivMol incorporates local contrastive molecular alignment without exposing raw molecules, labels, scaffolds, substructure masks, or embeddings. Experimental evaluation on widely used public molecular benchmarks, including ESOL, FreeSolv, Lipophilicity, BBBP, BACE, HIV, and Tox21, demonstrates that PrivMol provides a favorable trade-off among prediction accuracy, communication efficiency, empirical leakage resistance, and privacy protection. The study offers a practical route toward secure collaborative chemical intelligence for computer-aided drug discovery, toxicology prediction, and materials informatics. Full article

In semi-supervised learning scenarios, the presence of limited labeled data and abundant unlabeled samples poses significant challenges to model robustness and generalization. Although the semi-supervised broad learning system (SSBLS) effectively exploits manifold structure through graph Laplacian regularization, its optimization is typically formulated under the mean square error (MSE) criterion, which is sensitive to noise and outliers. To address this limitation, this paper introduces the maximum mixture correntropy criterion (MMC) into the SSBLS framework and proposes a model termed M2C-SSBLS. By replacing the conventional MSE loss with a mixture correntropy-based objective, the proposed method enhances robustness against non-Gaussian noise and abnormal samples while preserving the computational efficiency and analytical solution property of the BLS. Furthermore, to improve representation diversity and reduce model variance, a multi-view ensemble extension, named EC-SSBLS, is proposed. This method constructs multiple feature views through a random feature subspace strategy, and independently trains an M2C-SSBLS base learner on each subspace. Finally, the predicted results of each view are fused through a voting mechanism. Experiments on benchmark UCI datasets under noise-free, 10% and 20% label noise settings demonstrate that the proposed M2C-SSBLS consistently outperforms conventional SSBLS and other advanced semi-supervised learning approaches. The ensemble extension EC-SSBLS further enhances performance, particularly in noisy environments, validating the effectiveness of combining MMC-based optimization with multi-view ensemble learning. Full article

Deep-learning-based vulnerability detection has made substantial progress, but two limitations remain prominent. Sequence-based methods linearize source code and thus weaken the explicit modeling of control-flow and data-flow dependencies. Graph-based methods preserve program structure, yet conventional graph neural networks still have difficulty capturing long-range interactions in large code property graphs (CPGs). In addition, standard CPGs usually lack explicit variable semantics and security-critical node roles, which limits their ability to represent vulnerability-relevant program behavior. To address these issues, we propose SV-GEN, a vulnerability detection framework that combines large-language-model-driven semantic enhancement with hybrid sequence-graph learning. The novelty of SV-GEN lies in introducing a semantically enriched code property graph, termed Sem-CPG, which augments conventional CPGs with variable semantic roles and security-oriented node labels, and in coupling this representation with an adaptive fusion mechanism over structural and sequential views. Specifically, we use a large language model as an external semantic annotator to assign variable roles and identify source, sink, and sanitizer nodes, and then encode the resulting Sem-CPG with a Graph Transformer while modeling the code sequence with GraphCodeBERT. A learnable gating module is further used to adaptively fuse the graph-level and sequence-level representations for final prediction. Experiments on Devign, ReVeal, and DiverseVul show that SV-GEN achieves competitive or superior overall performance across benchmarks, with particularly strong improvements on the large and highly imbalanced DiverseVul dataset. Full article

The new energy vehicle (NEV) industry generates massive multi-source heterogeneous data. To overcome traditional database limitations in terminology disambiguation and multi-hop reasoning, this paper proposes a knowledge graph (KG)-based question-answering (QA) architecture. Three primary domain challenges are addressed: First, to tackle the poor semantic extraction of informal diagnostic texts, a deep semantic parsing network (BERT-BiLSTM-CRF) is integrated to extract high-precision knowledge from 150,000 real-world maintenance records. Second, to solve topological redundancy, the Labeled Property Graph (LPG) specification is employed to encapsulate parameters of 2157 vehicle models as internal attributes, significantly streamlining complex multi-hop reasoning. Finally, to enhance limited reasoning capabilities, an intent classification module (TextCNN) automatically translates natural language into graph queries, enabling deep fault tracing across up to five semantic levels. Experimental results demonstrate 98% and 93% accuracy in entity-relation recognition and intent classification, respectively. The resulting KG (8274 nodes, 14,488 edges) establishes a scalable paradigm for intelligent diagnostic reasoning in complex vertical domains. Full article

Background: Accurate identification of cancer driver genes is essential for understanding tumorigenesis and developing targeted therapies. Although graph neural networks (GNNs) have advanced multi-omics integration, existing methods often simply concatenate omics features and underutilize the topological information of biological networks. Methods: We propose M-GNN, a multi-modal GNN framework for cancer driver gene prediction. It employs separate Graph Convolutional Network (GCN) encoders to process four types of omics data (mutation, expression, methylation, copy number variation (CNV)), each represented as a 16-dimensional vector. We incorporate knowledge distillation by using soft labels from a pre-trained teacher model to enhance feature representation. An attention mechanism adaptively fuses the encoded omics features, and a dual-path classifier combining a GCN and a Multilayer Perceptron (MLP) preserves both intrinsic gene properties and network topology. Results: Experiments on three public protein–protein interaction (PPI) networks show that M-GNN consistently achieves the highest or second-highest AUPRC compared to five state-of-the-art methods. Ablation studies confirm the contribution of each module, and biological interpretability analysis—including analysis of GO enrichment and drug sensitivity—validates the reliability of the predicted genes. Conclusions: M-GNN provides a robust and interpretable computational tool for systematic cancer driver gene identification, effectively integrating multi-omics and network data. Full article

Let G be a graph with n vertices and m edges. A vertex magic total labeling of G is a bijection $f:V\left(G\right)\cup E\left(G\right)\to \{1,2,\dots ,n+m\}$ such that, for each vertex $u\in V\left(G\right)$, the sum of the label of u and the labels of all edges incident to u is equal to a fixed constant, referred to as the magic constant. A vertex magic total labeling is said to be even if the labels assigned to the vertices are exactly even numbers $\{2,4,6,\dots ,2n\}$. These labelings, along with related variations, have theoretical significance and practical applications, such as resource allocation, fault tolerance, and network design. Structured labelings aid channel assignment, address computation, and reduce collisions in networks. In this paper, we investigate wheel-related graphs that either admit or do not admit an even vertex magic total labeling. Furthermore, we introduce a new class of wheel-related graph, referred to as the plus wheel $W_n+r$, that can have such labelings, and we also establish a necessary and sufficient condition for such graphs to possess this property. Full article

Let $G=(V,E)$ be a finite simple graph with $p=\left|V\right|$ vertices and $q=\left|E\right|$ edges, without isolated vertices or isolated edges. A vertex magic total labeling is a bijection f from $V\cup E$ to the consecutive integers $1,2,\dots ,p+q,$ with the property that, for every vertex $u\in V$, one has $f\left(u\right)+{\sum }_{uv\in E}f\left(uv\right)=k$ for some magic constant k. The vertex magic total labeling is called E-super if furthermore $f\left(E\right)=\{1,2,\dots ,q\}$. A graph is called (E-super) vertex magic if it admits an (E-super) vertex magic total labeling. In this paper, we verify the existence of E-super vertex magic total labeling for a class of 3-regular graphs with a perfect matching, and we confirm the existence of such a labeling for general regular graphs of odd degree containing particular classes of 3-factors, which provides us with known and new examples. Note that Harary graphs are among the popular models used in communication networks. In 2012, G. Marimuthu and M. Balakrishnan raised a conjecture that if $n>4$, $n\equiv 0\left(mod4\right)$ and m is odd, then the Harary graph $H_{m,n}$ admits an E-super vertex magic labeling. Among others, we are able to verify this conjecture except for one case while $m=3$ and $n\equiv 4\left(mod8\right)$. Full article

Modelling connectivity in utility networks is essential for operational management, maintenance planning, and resilience analysis. The CityGML Utility Network Application Domain Extension (UNADE) provides a detailed conceptual framework for representing utility networks; however, most existing implementations rely on relational databases, where connectivity must be reconstructed through joins rather than represented as explicit relationships. This creates challenges when managing densely connected network structures. This study introduces the UNADE–Labelled Property Graph (UNADE-LPG) model, a graph-based representation that maps the classes, relationships, and constraints defined in the UNADE Unified Modelling Language (UML) schema into nodes, edges, and properties. A conversion pipeline is developed to generate UNADE-LPG instances directly from CityGML UNADE datasets encoded in GML, enabling the population of graph databases while maintaining semantic alignment with the original schema. The approach is demonstrated through two case studies: a schematic network and a real-world water system from Frankston, Melbourne. Validation procedures, covering structural checks, topological continuity, classification behaviour, and descriptive graph statistics, confirm that the resulting graph preserves the semantic structure of the UNADE schema and accurately represents the physical connectivity of the network. An analytical path-finding query is also implemented to illustrate how the UNADE-LPG structure supports practical network-analysis tasks, such as identifying connected pipeline sequences. Overall, the findings show that the UNADE-LPG model provides a clear, standards-aligned, and operationally practical foundation for representing utility networks within graph environments, supporting future integration into digital-twin and network-analytics applications. Full article

The Internet of Things (IoT) continues to expand across critical infrastructures, enabling automation, efficiency, and data driven decision making; yet, reliable device identification from network traffic remains hampered by severe class imbalance that skews learning and degrades performance. Synthetic data generation offers a promising remedy, particularly in privacy-sensitive security settings where access to representative traffic is limited. This paper advances the state of the art by proposing a framework that unites graph-conditioned generative modeling with large language model (LLM) guidance to produce realistic, semantically valid synthetic network traffic for imbalanced classification. First, we construct feature relationship graphs derived from Pearson correlation, Spearman rank correlation, and mutual information to capture inter-feature dependencies, and use these graphs to condition a Wasserstein GAN (WGAN), thereby preserving structural properties of real traffic during generation. Second, we employ an LLM to define class-specific semantic constraints, including admissible feature ranges, attribute correlations, and protocol level rules, which are enforced as soft guidance to steer the generator toward label-consistent and standards-compliant samples. Third, we institute a dual validation loop that combines LLM-based feedback on constraint satisfaction with evaluation of classifiers trained on datasets balanced by our method versus the traditional SMOTE technique. Lastly, extensive experiments demonstrate that jointly leveraging structural (graph) and semantic (LLM) conditioning yields higher-fidelity synthetic traffic and delivers consistent gains in macro-F1 and balanced accuracy for network traffic classification, highlighting the framework’s utility for security analytics under data scarcity and privacy constraints. Full article

Although ensemble clustering methods based on the co-association (CA) matrix have achieved considerable success, they still face the following challenges: (1) in the label space, the noise within the connective matrices and the structural differences between them are often neglected, and (2) the rich structural information inherent in the feature space is overlooked. Specifically, for each connective matrix, a symmetric error matrix is first introduced in the label space to characterize the noise. Then, a set of mapping models is designed, each of which processes a denoised connective matrix to recover a reliable consensus matrix. Moreover, multi-order graph structures are introduced into the feature space to enhance the expressiveness of the consensus matrix further. To preserve a clear cluster structure, a theoretical rank constraint with a block-diagonal enhancement property is imposed on the consensus matrix. Finally, spectral clustering is applied to the refined consensus matrix to obtain the final clustering result. Experimental results demonstrate that ECM-RCL achieves superior clustering performance compared to several state-of-the-art methods. Full article

Urban Digital Twins (UDTs) demand both simplified geometry and rich semantic information from Building Information Models (BIM) to be effectively integrated into Geospatial Information Systems (GIS). However, current BIM-to-GIS conversion methods struggle with geometric complexity and semantic loss, particularly at scale. This paper proposes a novel, scalable methodology for comprehensive BIM/GIS integration, addressing both geometric and semantic challenges. The approach introduces a geometry conversion workflow that transforms solid BIMs into valid, simplified CityGML representations through a level-by-level detection of building elements and outer surface extraction. To preserve semantic richness, all entities, attributes, and relationships—including implicit connections—are automatically extracted and stored in a Labeled Property Graph (LPG) database. The method is further extended with a new CityGML Application Domain Extension (ADE) that supports Multi-LoD4 representations, enabling selective interior visualization and efficient rendering. A web-based urban digital twin platform demonstrates the integration, allowing dynamic semantic querying and scalable 3D visualization. Results show a significant reduction in geometric complexity, full semantic retention, and robust performance in visualization and querying, offering a practical pathway for advanced UDT development. Full article

Molecular odor prediction represents a fundamental challenge in computational chemistry with significant applications in fragrance design, food science, and chemical safety assessment. While traditional Quantitative Structure–Activity Relationship (QSAR) methods rely on hand-crafted molecular descriptors, recent advances in graph neural networks (GNNs) enable direct end-to-end learning from molecular graph structures. However, systematic comparison between these approaches for multi-label odor prediction remains limited. This study presents a comprehensive evaluation of traditional QSAR methods compared with modern GNN approaches for multi-label molecular odor prediction. Using the GoodScent dataset containing 3304 molecules with six high-frequency odor types (fruity, green, sweet, floral, woody, herbal), we systematically evaluate 23 model configurations across traditional machine learning algorithms (Random Forest, SVM, GBDT, MLP, XGBoost, LightGBM) with three feature-processing strategies and three GNN architectures (GCN, GAT, NNConv). The results demonstrate that GNN models achieve significantly superior performance, with GCN achieving the highest macro F1-score of 0.5193 compared to 0.4766 for the best traditional method (MLP with basic preprocessing), representing a 24.1% relative improvement. Critically, we discover that threshold optimization is essential for multi-label chemical classification. These findings establish GNNs as the preferred approach for molecular property prediction tasks and provide crucial insights for handling class imbalance in chemical informatics applications. Full article

Green construction in infrastructure projects has emerged as a crucial approach for reducing environmental impacts, yet its implementation is fraught with numerous uncertainties. To assess the capacity to maintain and restore green construction performance in complex environments, this study proposes the concept of Green Construction Performance Resilience (GCPR) for infrastructure projects and develops methodologies for its management and optimization. This study constructs a project network based on the labeled property graph (LPG) technique and demonstrates its dynamic evolution throughout the entire project lifecycle. A series of indicators for quantifying GCPR are constructed and applied, enabling the quantification of green construction performance resilience in infrastructure projects. An optimization method for GCPR based on genetic algorithms is proposed. Finally, the applicability and effectiveness of the proposed methodologies are validated through the analysis of real-world infrastructure project cases. The results demonstrate that the project network model can comprehensively capture the complexity of large-scale infrastructure projects, and that the GCPR indicators effectively measure green construction performance resilience, providing valuable decision-making support for project managers. The optimization algorithm has been validated and shown to improve the GCPR level of the project. This study enriches interdisciplinary research on project resilience and project complexity theory and provides project managers with quantitative analysis and visualization tools to facilitate the attainment of green construction performance objectives in infrastructure projects and accelerate the transition towards low-carbon practices. Full article

Graph neural networks (GNNs) have exhibited remarkable performance in various applications. Still, research has revealed their vulnerability to backdoor attacks, where Adversaries inject malicious patterns during the training phase to establish a relationship between backdoor patterns and a specific target label, thereby manipulating the behavior of poisoned GNNs. The inherent symmetry present in the behavior of GNNs can be leveraged to strengthen the robustness of GNNs. This paper presents a quantitative metric, termed Logit Margin Rate (LMR), for analyzing the symmetric properties of the output landscapes across GNN layers. Additionally, a learning paradigm of graph self-distillation is combined with LMR to distill the symmetry knowledge from shallow layers, which can serve as the defensive supervision signals to preserve the benign symmetric relationships in deep layers, thus improving both model stability and adversarial robustness. Experiments were conducted on four benchmark datasets to evaluate the robustness of the proposed Graph Self-Distillation-based Backdoor Defense (GSD-BD) method against three widely used backdoor attack algorithms, demonstrating the robustness of GSD-BD even under severe infection scenarios. Full article

This paper introduces a novel framework enabling natural language question answering on Piping and Instrumentation Diagrams (P&IDs), addressing a critical gap between engineering design documentation and intuitive information retrieval. Our approach transforms static P&IDs into queryable knowledge bases through a three-stage pipeline. First, we recognize entities in a P&ID image and organize their relationships to form a base entity graph. Second, this entity graph is converted into a Labeled Property Graph (LPG), enriched with semantic attributes for nodes and edges. Third, a Large Language Model (LLM)-based information retrieval system translates a user query into a graph query language (Cypher) and retrieves the answer by executing it on LPG. For our experiments, we augmented a publicly available P&ID image dataset with our novel PIDQA dataset, which comprises 64,000 question–answer pairs spanning four categories: (I) simple counting, (II) spatial counting, (III) spatial connections, and (IV) value-based questions. Our experiments (using gpt-3.5-turbo) demonstrate that grounding the LLM with dynamic few-shot sampling robustly elevates accuracy by 10.6–43.5% over schema contextualization alone, even under high lexical diversity conditions (e.g., paraphrasing, ambiguity). By reducing barriers in retrieving P&ID data, this work advances human–AI collaboration for industrial workflows in design validation and safety audits. Full article