Search Results (728)

In this work, we introduce and study a generalization for r-uniform hypergraphs of complement-reducible graphs, the so-called co-graphs. The operations for r-join-hypergraphs are the binary disjoint union of two given r-join-hypergraphs and the r-nary join, which inserts all possible hyperedges of cardinality r, each including exactly one vertex from each of r given r-join-hypergraphs. We characterize the primal graphs of r-join-hypergraphs as special co-graphs and give some properties of r-join-hypergraphs. This allows us to give a method that decides whether an r-uniform input hypergraph H is an r-join-hypergraph and, in the case of an affirmative answer, finds a decomposition tree for H in polynomial time. This task proved to be challenging. We show specific formulas for computing various hypergraph parameters for r-uniform hypergraphs defined by the binary disjoint union of two r-uniform hypergraphs and the r-nary join of rr-uniform hypergraphs. The parameters considered are the size of a largest stable set, the size of a largest co-stable set, the size of a largest independent set, the size of a largest co-independent set, the size of a smallest vertex cover, the size of a smallest 2-transversal, the size of a largest matching, the size of a smallest dominating set, the chromatic number, the strong chromatic number, and the upper chromatic number. This yields $\mathcal{O}(n·r)$-time algorithms to compute these values on r-join-hypergraphs on n vertices given by a decomposition tree. Of particular interest is the development of an efficient algorithm to compute the size of a largest matching. In addition, we infer relations among the considered parameters restricted to r-join-hypergraphs. Our methods generalize and reprove a number of results that are known for co-graphs. Full article

In the digital transformation of the petroleum industry, massive multi-source heterogeneous tables are distributed across databases, Word documents, PDF reports, and engineering systems. Their unstructured format and weakly expressed inter-table relationships make it difficult for conventional keyword-based or single-table retrieval methods to locate the complete set of tables needed for complex queries. To address this problem, this paper proposes Relatab, a multi-table retrieval framework based on implicit association reasoning. Relatab first estimates query–table relevance through a dual-level semantic matching mechanism that combines table-level signals, including captions and column names, with value-level signals weighted by entropy and CRITIC criteria. It then constructs an implicit table graph using column-name and column-content similarity, and applies a max-product multi-hop propagation rule with decay and pruning to identify complementary tables that are not directly matched by the query. Finally, direct relevance and inter-table complementarity are fused to produce the retrieved table set. Experiments on Spider, Bird, and CementingTables show that Relatab achieves Top-2 recall rates of 79.21%, 61.26%, and 77.88%, respectively, outperforming DTR by 1.74, 2.33, and 1.85 percentage points. The results indicate that explicit modeling of implicit inter-table associations improves retrieval coverage in complex multi-table scenarios while remaining applicable to petroleum-domain documents. Full article

Retrieval-augmented generation (RAG) has been widely used to improve factual grounding in small language models (SLMs). However, many existing retrieval methods rely mainly on shallow semantic similarity between queries and text chunks. As a result, the retrieved evidence is often fragmented or weakly connected, which can limit multi-hop reasoning and sometimes lead to hallucinated answers. In this work, we propose Hallucination-Aware Multi-Objective Ant Colony Optimization (HA-MOACO), a structure-aware Graph-RAG framework that models relationships between pieces of evidence and searches for useful reasoning paths in a graph structure. The framework first constructs a domain knowledge graph and identifies several candidate entry nodes related to the query. Starting from these nodes, an ant colony optimization (ACO) strategy is used to explore possible evidence paths. During this process, multiple signals are considered simultaneously, including semantic relevance, structural connectivity between nodes, and reliability indicators that help reduce contradictory or low-confidence evidence. The selected evidence paths are then used to build a compact and grounded context for the language model. This process helps keep important reasoning connections while reducing the influence of irrelevant information. Experimental results on reasoning-focused benchmarks show that the proposed framework improves answer accuracy by about 6%, increases F1 by roughly 5–6%, and improves exact-match scores by around 7% compared with strong graph-based retrieval baselines. At the same time, the hallucination rate is reduced by about 40%, while retrieval latency is roughly 20% lower. These results suggest that combining structure-aware retrieval with reliability-oriented optimization can improve both reasoning quality and factual consistency in SLMs, while still remaining efficient for practical deployment. Full article

Reliable wearable activity recognition requires not only a class label but also an auditable indication of whether that label is supported by historical sensor evidence. We present CC-NSIEA, a label-preserving neural-plus-rule-based class-contrast evidence audit for cross-subject wearable activity recognition. A Temporal Residual Perception Network supplies the sole activity label, posterior probabilities, and a normalized temporal embedding. A read-only Training-Subject Evidence Memory retrieves global, predicted-class, and competing-class records. A rule-based Evidence Consistency Audit combines data validity, dynamic/static motion coherence, retrieval support, and class separation. When first-round evidence is insufficient, Class-Contrast Evidence Refinement performs one deterministic contrast between the predicted class and the strongest posterior competitor; the audit cannot change the neural label. The term neuro-symbolic is used only in this restricted architectural sense: a neural predictor is coupled to explicitly represent deterministic predicates and a finite rule-based controller; the method does not perform symbolic inference, theorem proving, or knowledge-graph reasoning. On five subject-disjoint outer folds of the UCI HAR official training partition, the shared perception model achieved 90.13% accuracy and 90.55% macro-F1 across 7352 out-of-fold windows from 21 subjects. Relative to a matched dynamic deterministic controller, CC-NSIEA increased Error AUPRC from 0.423802 to 0.433057 and reduced AURC from 0.035941 to 0.035913. The 10,000-resample subject-cluster bootstrap interval for the AUPRC difference was [0.001595, 0.019547]. CC-NSIEA provides an evidence-centered complement to confidence-based reliability estimation. Full article

Human Digital Twins (HDTs) rely on reliable context estimation to support personalized services, adaptive decision-making, and context-aware interaction, yet local ego sensing can be noisy, intermittent, or ambiguous. This study investigates whether Social Internet of Things (SIoT)-enabled opportunistic recruitment of external sources can improve HDT context estimation under partial observability. A controlled synthetic simulator is developed in which the SIoT layer is represented as a typed object graph supporting graph-constrained candidate discovery, bounded relationship-guided discovery, and cost-aware recruitment. The evaluation compares ego-only sensing, a high-coverage opportunistic-all reference, SIoT-aware bounded recruitment, and privacy-aware SIoT recruitment across nominal conditions, degraded ego sensing, ambiguous local context, and noisy/untrusted external sources. Performance is assessed with strict joint overall context accuracy, mean variable accuracy, per-variable Macro-F1, operational cost, effective recovery cost, source-cap-matched baselines, discovery-mode comparisons, and graph-size scalability. The results show that opportunistic-all recruitment gives the highest raw OCA because it recruits many sources, whereas bounded SIoT-aware policies provide lower-cost operating points and preserve nearly the same accuracy as exhaustive SIoT discovery in the discovery-mode comparison. The findings are therefore framed as accuracy–cost–privacy–scalability trade-offs in a synthetic, reproducible testbed rather than as deployment-level claims of universal policy superiority. Full article

Retrieval-Augmented Generation (RAG) grounds large language models in external evidence, but many RAG systems represent knowledge either as flat text chunks or as automatically constructed indexing graphs. This assumption is incomplete for curated thematic corpora, including religious scriptures, legal codes, clinical guidelines, educational taxonomies, policy documents, and library classification systems, where domain experts have already organized knowledge into thematic paths and citeable canonical units. This paper investigates how RAG can exploit such expert-authored structures while pruning evidence to a compact and query-specific set. We conduct a critical survey supported by a bibliometric analysis of 2815 Scopus-indexed RAG-related records exported on 26 May 2026, of which 2809 records were retained after duplicate removal. The bibliometric results indicate rapid growth in RAG research but limited explicit consolidation around curated thematic paths, canonical evidence units, or thematic path-guided evidence pruning. We therefore propose ThemePath-RAG, a retrieval framework that retrieves curated thematic paths as high-recall semantic routes, expands candidate canonical evidence, and applies query-aware scoring and global pruning before generation. To assess operational feasibility, we implement ThemePath-RAG for Qur’anic question answering and compare it with a Vector RAG baseline on 150 paired questions using RAGAS context relevance with gpt-4o-mini as the LLM evaluator. Both methods return approximately three final ayat per question. Vector RAG achieves higher mean context relevance than ThemePath-RAG ($0.920$ versus $0.798$; $p<0.001$). Thus, the proof of concept establishes the feasibility of thematic-path-guided retrieval and identifies evidence-selection challenges, rather than demonstrating superiority over conventional vector retrieval. The paper clarifies the framework’s relationship to GraphRAG, LightRAG, HippoRAG, PathRAG, ontology-based RAG, and AI-augmented bibliometric systems, and outlines a language-matched, multi-baseline evaluation agenda for future cross-domain validation. Full article

Designing sustainable closed-loop supply chain (CLSC) networks requires jointly assessing node-level operational attributes (recovery efficiency, processing capacity, unit cost) and inter-node spatial structure. Existing methods, including mixed-integer programming, multi-objective metaheuristics, and graph-matching, typically optimize a single cost dimension and do not decompose structural connectivity from attribute-level inefficiency. We propose a Fused Gromov–Wasserstein (FGW) diagnostic framework that combines the Wasserstein distance (attribute similarity) and the Gromov–Wasserstein distance (structural alignment) via a convex trade-off parameter α, solved using the conditional gradient algorithm. Supply–capacity imbalances are resolved by marginal rescaling, with residual unabsorbed mass reported as a diagnostic indicator of infrastructure shortfall. The framework is applied to an eight-echelon PET bottle recovery and filament manufacturing network across 24 synthetic benchmark instances at three scale classes. The FGW cost decomposes exactly into feature and structural components, allowing bottleneck arcs to be diagnosed as attribute-driven or structure-driven. Under this benchmark, bottleneck cost decreases with network size, the most frequent bottleneck arc shifts from the collection interface in small networks to the mid-chain processing handoff in large networks, and attribute heterogeneity accounts for the majority of FGW cost (57.9%, conditional on the normalization and weighting scheme used) across all 144 arc–instance combinations. These results position FGW as a tractable, interpretable diagnostic layer for circular supply chain analysis, complementing rather than replacing classical CLSC design models. Full article

Targeting the joint challenges posed by sparse graph topology, limited semantic expressiveness, and scarce annotation resources that commonly afflict knowledge graphs in the oil and gas pipeline safety domain, this paper presents a Multi-Information Fusion Unsupervised Entity Alignment model (MIF-UEA). The proposed method constructs high-quality initial alignment pairs by integrating multi-source similarity computation with a structure-aware seed generation mechanism and performs representation learning by fusing structural features and semantic attribute information. Furthermore, a pseudo-label augmentation and denoising strategy is introduced to enhance the effectiveness of self-training. Finally, entity matching is achieved through an optimal transport model. Experimental results confirm that MIF-UEA surpasses existing baselines across both the specialized oil and gas pipeline safety dataset and multiple general-domain benchmarks, demonstrating its effectiveness and generalization capability. Full article

Accurate localization is a critical requirement for autonomous vehicle (AV) navigation, particularly in environments where GPS signals are unreliable or unavailable. A wide range of LiDAR-based point cloud similarity metrics have been proposed for high-definition (HD) map localization, but systematic comparisons of distinct metric families on the same real-world dataset, under identical conditions, remain scarce. In this paper, we present an offline comparative study of three point cloud similarity metrics within a unified HD map-based localization framework, under the assumption of largely static environments and planar, yaw-dominant vehicle motion typical of on-road driving. The HD map is constructed as a directed graph of GPS coordinates, each linked to a corresponding LiDAR scan, collected over a 30-min drive on a university campus using a Velodyne VLP-16 sensor and a ublox ZED-F9P RTK-GPS receiver, yielding 19,500 time-synchronized point clouds. Within this framework, we develop and compare three similarity metrics drawn from distinct families: Fast Point Feature Histograms (FPFH) with KDTree-based matching, Procrustes-based alignment via singular value decomposition, and a planar projection method based on 2D angular histogram cross-correlation. Each metric is evaluated on the same dataset in terms of similarity score profile (localizability) and per-pair computational cost. FPFH provides rich local geometric matching but at an average per-pair cost of approximately 1018 s, making it suitable only for offline analysis. Procrustes alignment yields the smoothest score profiles, with an exact self-similarity baseline of zero, at an average of $2.63$ s per pair. The planar projection method produces the most location-invariant profiles at an average of $11.6$ s per pair. We also discuss the recursive localization architecture into which any of these metrics could be embedded, and analyze the gap between current per-pair costs and what would be required for online deployment, which we identify as a direction for future work. The study contributes a controlled, reproducible benchmark of three metric families on a single real-world dataset, and provides guidance for selecting similarity metrics under stated operating assumptions. Full article

Career growth path planning is still dominated by statistical association models that summarize historical transitions but do not explicitly represent the causal mechanisms linking capability development, project exposure, policy support, performance improvement, and promotion outcomes. This study develops a reproducible simulation benchmark for evaluating whether prompt-structured priors, when coupled with dual validation, can help assemble intervention-ready career causal graphs. A structural causal model (SCM) first generated 20,000 synthetic career trajectories with known ground-truth dependencies among ten variables, including education, experience, training hours, certification, project exposure, performance, and promotion. Four prompt families-zero-shot, few-shot, Chain-of-Thought (CoT), and CoT plus schema constraints-were instantiated through a controlled prompt-response emulator so that prompt structure could be studied independently of vendor-specific model drift. The emulator gradients should therefore be read as literature-informed design assumptions about structured prompting rather than as empirical measurements from any named production LLM. Candidate edges were subsequently refined by data validation and expert-proxy domain rules. In the main 30-run benchmark, the best prompt-only setting (CoT plus schema) achieved an F1-score of 0.842, while the proposed hybrid method achieved an F1-score of 0.959 and an intervention-effect mean absolute error of 0.0046. Run-wise confidence intervals and approximate significance checks further indicated that the hybrid workflow materially outperformed the prompt-only variants under the benchmark protocol. A public employee-promotion dataset (N= 54,808) was further used as an external plausibility anchor, where KPI attainment, awards, previous ratings, training score, and length of service were all positively associated with promotion. The results indicate that prompt-structured priors can be useful as a transparent proposal-and-validation mechanism, but not as a substitute for direct validation on real LLMs, matched comparisons with standard causal-discovery baselines, or real HR deployment settings. Accordingly, the central aim is a domain-specific methodological benchmark for testing prompt-structured proposal mechanisms in career-growth causal modeling, rather than a claim of standalone LLM causal discovery or a universal benchmark for every causal-discovery setting. Full article

Images captured by vision sensors provide visual evidence for scene understanding, including object appearances, pixel-level regions, and spatial relations among entities. Panoptic Scene Graph Generation (PSG) constructs structured scene representations by grounding visual entities with panoptic masks and predicting relationships among objects and regions. In pair-then-relation PSG pipelines, subject–object pair recall is critical to final triplet recall. However, existing pair proposal approaches mainly score candidate subject–object pairs based on object–query feature matching, while mask-derived spatial cues such as object locations, relative geometry, and local layouts remain underexplored. Consequently, ground-truth subject–object pairs may be excluded from the Top-$K_r$ proposals before relation decoding. To address this problem, this paper proposes a Spatially Aware Pair Proposal Model (SAPPM), which incorporates mask-derived soft centroids, relative geometry, and local-neighborhood context into pair scoring. SAPPM uses Grouped Vector Attention (GVA) to model local spatial interactions and introduces a spatially adaptive gating module to calibrate spatial-branch contributions. Experiments on the PSG dataset under the Scene Graph Detection (SGDet) protocol show that SAPPM achieves competitive performance, reaching 32.53 R@20 and 27.36 mR@20. These results indicate that SAPPM improves PSG performance by enhancing ground-truth pair coverage in the candidate proposal set. Full article

To address the problem of high-frequency roll disturbances generated during dynamic balancing in non-coaxial two-wheeled robots, this paper proposes a Rotation-Optimized Inertial–Visual SLAM system (ROIV-SLAM) for robust state estimation. The proposed approach adopts a decoupled architecture for translation and rotation estimation. In the front-end, an Extended Kalman Filter (EKF) is employed to fuse LiDAR, an inertial measurement unit (IMU), and wheel odometry to obtain an initial translation estimate. Meanwhile, a physical manifold constraint is constructed using the gravity vector and surface normals extracted from RGB-D point clouds, supporting stable rotation estimation under high-frequency disturbances through Lie-group-based optimization. In the back-end, a factor graph is established, and loop closure robustness is enhanced through vision–LiDAR scan matching. Experimental results indicate that ROIV-SLAM achieves improved trajectory consistency with respect to the optimized reference trajectory and more robust mapping performance compared with the evaluated baseline approaches in the tested scenarios. The results further suggest that introducing task-specific physical dynamic constraints and a decoupled estimation mechanism helps suppress high-frequency motion noise inherent to balancing robots, thereby improving the robustness of state estimation in complex environments. Full article

Explicit formulas for graphical Stirling and Bell numbers are known for relatively few graph families. We derive exact expressions for three classes whose independence structure admits a complete combinatorial description: complete multipartite graphs, the graph obtained from a balanced complete bipartite graph by deleting a perfect matching, and the Mycielskian of a star. For complete multipartite graphs we express the graphical Stirling number as a convolution of classical Stirling numbers across the partite classes, and we recover the known factorization of the graphical Bell number as a product of classical Bell numbers. For the matching-deleted graph we show that its graphical Bell number is a binomial convolution of squared Bell numbers, which we identify as a moment of a product of two independent Poisson random variables with unit mean. This representation yields log-convexity of the sequence, a sharp exponential lower bound, a two-sided estimate, and a Laplace-transform identity. For the Mycielskian of a star, a decomposition according to the block containing the original center vertex, together with Vandermonde’s convolution and a Stirling recurrence, gives a single-sum closed form for the graphical Stirling numbers, from which two explicit evaluations follow. Several resulting integer sequences appear in the OEIS, and one Bell-number sequence appears not to be currently recorded there. Full article

Person re-identification (re-ID) aims to match pedestrian images across disjoint camera views. Existing multi-source unsupervised domain adaptation (UDA) re-ID methods still face two critical issues: they fail to effectively balance domain-invariant feature learning and domain-specific style preservation and cannot adequately model the implicit correlations among diverse source domains, resulting in limited cross-domain generalization performance. To address these challenges, this paper proposes a novel multi-source UDA re-ID framework equipped with a Mixture of Experts feature extraction (MEFE) network and a Graph-Based Relation (GBR) module. Specifically, the MEFE network integrates mixed Instance and Batch Normalization (MIBN) to extract robust domain-invariant features, while the embedded domain-specific style information (DSI) module compensates for lost domain-specific style details at the feature level. Furthermore, the cascaded Graph Attention and Graph Convolution Networks (GATs/GCNs) in the GBR module adaptively explore implicit feature correlations and achieve effective multi-source feature fusion. Center maximum mean discrepancy loss is adopted to further reduce cross-domain distribution discrepancies. Extensive experiments on large-scale datasets demonstrate that the proposed method achieves state-of-the-art performance and substantially outperforms mainstream UDA re-ID approaches. Full article

Safety signs in innovative manufacturing environments fail to match dynamic risks due to the separation of perception, semantics, and decision-making. Existing methods lack closed-loop integration of IoT sensor streams, knowledge graph reasoning, and adaptive signage control. This paper proposes a framework that fuses dynamic graph attention networks with hierarchical temporal knowledge graphs and reinforcement learning optimization. The framework extracts spatiotemporal dependencies from multi-source sensors, traces risk propagation paths on an industrial knowledge graph, and generates adaptive signage actions. Experimental results demonstrate that the proposed method achieves 96.7% risk identification accuracy, a 91.3% risk propagation F1 score, a 94.2 semantic matching score, and 43.65 milliseconds response latency. Real-world validation on an aerospace workshop confirms the method’s effectiveness. This work provides a closed-loop solution from physical perception to adaptive semantic expression for intelligent manufacturing safety. Full article