Search Results (12,229)

Rare and safety-critical traffic situations remain challenging for autonomous driving (AD) because they are underrepresented in common training data and may include objects outside standard detector classes. This paper presents a real-time RGB-LiDAR fusion framework for detecting and reacting to rare traffic situations in CARLA (Car Learning to Act), a reproducible simulator for AD research. The approach combines YOLOv8n-based RGB perception, bird’s-eye-view (BEV) LiDAR clustering, decision-level fusion, an interpretable rule-based safety agent with hysteresis, Time-to-Collision (TTC)-aware escalation, and an automatic emergency braking (AEB) override above the CARLA autopilot. Fused observations are classified as semantic–geometric detections, semantic-only detections, or geometric-only obstacle candidates, where unmatched LiDAR clusters are treated conservatively as candidate-level physical evidence rather than confirmed rare objects. The framework was evaluated on three CARLA maps and 3CSim-inspired corner-case scenarios comprising 19,253 frames, with additional weather/lighting stress tests and a public nuScenes mini cross-platform check. On a manually annotated subset of 4800 CARLA frames, corresponding to approximately $24.9\%$ of the recorded CARLA log, the full framework achieved $96.2\%$ precision, $97.3\%$ recall, and a $96.7\%$ F1-score for safety-relevant threat detection. The control experiments show that the fusion-based safety agent reduced unnecessary braking to $1.7\%$ compared with $8.6\%$ for the LiDAR-only baseline and achieved event-level success on the annotated critical intervals. The proposed CPU-only implementation maintained real-time performance, with an average processing time of $34.7\mathrm{ms}$. Full article

We characterize observability in representation learning through Platonic Projection Structures (PPS), an operator-theoretic framework for analyzing representation accessibility under partial observation. Rather than treating observable outputs as direct reflections of latent representations, PPS models observation as a geometry induced by a self-adjoint positive semidefinite operator acting on a latent Hilbert space. A system is represented as a triple $(\mathcal{H},\Pi ,O),$ where $\mathcal{H}$ denotes a latent representation space, $\Pi ⪰0$ is an observation operator, and $O\left(v\right)=⟨v,\Pi v⟩$ defines an induced scalar observable. The framework characterizes observability through the quotient geometry $\mathcal{H}/ker\left(\Pi \right),$ which represents equivalence classes of latent states that are indistinguishable under observation. From this perspective, observable behavior is governed not by latent representations themselves, but by the geometry induced through the observation operator. We show that both quantum measurement and representation inference under linear observation models can be formulated within this common operator-theoretic structure while differing in the algebraic properties of their observation operators. Within this perspective, quantum measurement serves primarily as a mathematically canonical example of projection-mediated observability. The correspondence developed in PPS is therefore structural rather than physical. Within the same framework, representation transfer and knowledge distillation can be interpreted as approximate preservation of observable geometry through the intertwining condition $\Phi {\Pi }_T\approx {\Pi }_S\Phi .$ PPS further reveals a structural limitation of output-based interpretability: latent components contained in $ker\left(\Pi \right)$ are fundamentally inaccessible from observables generated through the induced observation process. Accordingly, attribution and explanation methods inherit intrinsic constraints imposed by the observation geometry itself. We provide controlled empirical validations demonstrating kernel-invariant observability, projection-induced attribution gaps, and rank-controlled observable geometry in latent representation spaces. Overall, PPS provides a mathematically explicit characterization of observability through operator-induced quotient geometry, offering a unified perspective on representation accessibility, interpretability, and representation transfer. Full article

In this study, we present a novel adaptive algorithm, social learning-enhanced deep reinforcement learning (SLDRL), which integrates social learning mechanisms into deep reinforcement learning (DRL) to improve agent performance in both discrete and continuous state-space environments. The proposed hybrid control architecture enables agents to autonomously decide when and how to exploit socially acquired behaviors, balancing social learning with individual exploration through an entropy-based intrinsic motivation mechanism. The framework incorporates online imitation and enactment mechanisms that allow agents to observe and selectively reuse behavioral sequences acquired from other agents during training. We evaluate SLDRL in a sparse-reward discrete grid-based foraging task and in the dense-reward continuous-state/discrete-action CartPole problem. In both domains, SLDRL agents outperform baseline DRL agents, achieving faster learning and higher cumulative rewards. The results show that socially acquired behaviors are utilized adaptively throughout training, with the balance between imitation and individual learning emerging dynamically according to the structure of the environment and the agent’s experience. Comparisons with a behavioral cloning baseline further indicate that selectively integrating observed behaviors can yield more robust long-term learning than direct imitation of demonstration trajectories. Overall, the results demonstrate that SLDRL can effectively leverage online social learning in diverse environments. Full article

Feature importance analysis is essential for interpreting machine learning models in diabetes mellitus (DM) risk prediction; however, existing interpretability methods often produce inconsistent feature rankings across models. This study proposes a unified ODE-inspired interpretability framework and an algorithmic decision procedure for robust feature selection by integrating contribution-based (SHAP), perturbation-based (permutation importance), and sensitivity-based feature importance measures. Multiple supervised machine learning models, including Logistic Regression, Random Forest, Gradient Boosting, Histogram Gradient Boosting, and Multilayer Perceptron, were trained and evaluated on a longitudinal biochemical and demographic dataset comprising 200 patients with three repeated visits (N = 600 observations). To preserve longitudinal integrity and avoid patient-level information leakage, grouped cross-validation was applied. A sensitivity-based feature importance formulation using finite-difference approximations enabled model-agnostic comparison across heterogeneous machine learning architectures. Stability, normalization, and cross-method agreement analyses were additionally introduced to evaluate consistency of feature rankings across models and interpretability methods. Experimental results consistently identified HbA1c as the dominant predictor, followed by lipid-related variables, age, and body mass index. Strong agreement was observed between ODE-inspired feature importance and SHAP analysis, whereas permutation importance demonstrated comparatively weaker agreement with sensitivity-based methods. The proposed framework further enabled systematic analysis of ranking stability, cross-method agreement, longitudinal sensitivity dynamics, and the introduction of an agreement-weighted Consensus Interpretability Score (CIS) for unified feature ranking across heterogeneous interpretability methods. The results demonstrate that integrating ODE-inspired sensitivity analysis with machine learning provides a robust, interpretable, and computationally scalable framework for feature importance assessment in diabetes risk prediction. The proposed approach offers a principled solution to inconsistent feature importance estimation and supports more reliable interpretation of biomedical machine learning models. Full article

Forecasting fine particulate matter with a diameter of 2.5 μm (PM_2.5) is critically important due to its adverse effects on human health and environmental sustainability. Although ground-based monitoring stations provide accurate measurements, their limited spatial coverage restricts large-scale PM_2.5 assessment, especially in complex urban regions. Consequently, aerosol optical depth (AOD) derived from satellite imagery, combined with advanced deep learning (DL) techniques, has emerged as an effective alternative by offering wide spatial coverage and rich spatio-temporal information. This paper proposed an optimization-driven fuzzy transformer–deep belief network (ODFT-DBN) for accurate PM_2.5 air pollution prediction. The proposed framework integrates a fuzzy inference module to model uncertainty and nonlinear environmental relationships, a transformer encoder to capture long-range spatio-temporal dependencies, and a DBN to extract hierarchical features and improve prediction robustness. In addition, a novel multi-objective gray wolf optimizer (NMOGWO) is employed to jointly optimize the model hyper-parameters and fuzzy membership functions. The proposed approach is implemented for the city of Tehran, Iran, using meteorological variables, topographical features, ground-based PM_2.5 measurements, and satellite-derived AOD data. The ODFT-DBN model is compared with several benchmark methods, including bidirectional encoder representations from transformers (BERT), transformer, long short-term memory (LSTM), gated recurrent unit (GRU), convolutional neural network (CNN), DBN, and extreme gradient boosting (XGBoost). Experimental results demonstrate that the proposed framework achieves superior predictive performance, attaining an R² value of 0.94 and root mean square error (RMSE) of 0.8 μg/m³. Scatter plot analyses indicate a strong agreement between predicted and observed PM_2.5 values, while the proposed model exhibits low variance, stable convergence behavior, and acceptable computational time. Overall, the results confirm the effectiveness, robustness, and practical applicability of the proposed ODFT-DBN framework for spatio-temporal PM_2.5 forecasting. Full article

As a critical alpine transition zone linking the Qinghai–Tibet Plateau and the Loess Plateau, Gannan Prefecture acts as an important water conservation area in the upper Yellow River basin of China. Based on GLDAS-2.1 surface soil moisture (SSM) datasets spanning 2000–2022 and interpretable machine learning tools (SHAP and ALE), this paper analyzes the spatiotemporal evolution, future trend sustainability, and nonlinear statistical associations between environmental predictors and SSM. The main results were as follows: (1) SSM exhibited a significant upward trend with an annual growth rate of 0.18 kg·m⁻²·a⁻¹ (p < 0.001), and an abrupt turning point occurred in 2017. The spatial pattern of high SSM in the southeast and low SSM in the northwest remained relatively stable, with the centroid migration distance being less than 1.81 km; most regions presented statistically significant moistening trends (p < 0.05). (2) Natural environmental predictors jointly carried 95.79% of the total statistical explanatory weight for modeled SSM variability. Precipitation possessed the highest explanatory proportion (37.93%), followed by temperature (27.30%), potential evapotranspiration (ET_p, 12.26%), elevation (10.44%), and fractional vegetation cover (FVC, 7.77%). One-dimensional ALE curves identified sample-limited statistical breakpoints: SSM gradually plateaued when precipitation reached 650–700 mm, while modeled SSM decreased substantially once ET_p exceeded 800 mm·a⁻¹. Two-dimensional ALE further characterized combined statistical correlations among precipitation, temperature, and ET_p. Model outputs also indicated that FVC above 0.45 corresponded to enhanced soil water retention within the observed sample range, which only reflects statistical patterns captured in this dataset rather than universal regulatory standards. This study offers quantitative statistical understanding of SSM variations across alpine transition zones. Full article

Objectives: To assess the feasibility and reproducibility of predicting antimicrobial resistance (AMR) in Escherichia coli from MALDI-TOF mass spectrometry data using a standardized, open-source machine learning (ML) workflow, we systematically compared four ML algorithms, evaluated the impact of culture conditions, extract storage, and spectral preprocessing on model performance, and validated results through nested cross-validation with statistical significance testing. Methods: A total of 282 clinical E. coli isolates were analyzed. Two MALDI-TOF MS datasets were generated from freshly cultured extracts (T₁) and recultured isolates one year later (T₃), yielding 4468 spectra. A third dataset from the T₁ extracts stored at −20 °C for one year (T₂) was evaluated for spectral stability but excluded from primary modeling likely due to storage-induced degradation. Protein spectra (m/z 2000–15,000) were preprocessed using an in-house developed MALDI-TOF preprocessing pipeline (MTPP) comprising variance stabilization, Savitzky–Golay smoothing, SNIP baseline correction, TIC normalization, LOWESS alignment, and MAD-based peak detection (SNR ≥ 3), yielding 121 m/z features. Four classifiers—Random Forest (RF), Logistic Regression, Support Vector Machine, and Gradient Boosting—were trained to predict resistance to 11 antibiotics using nested cross-validation: outer GroupShuffleSplit (5-fold, isolate-level) for evaluation and inner GroupKFold for recursive feature elimination (RFECV) and hyperparameter tuning (RandomizedSearchCV). Classification thresholds were optimized via the precision–recall curve. Model performance was assessed using AUROC, AUPRC, F1-score, Matthews Correlation Coefficient (MCC), and bootstrap 95% confidence intervals (1000 replicates). Pairwise model comparisons were tested with McNemar’s chi-squared test. Results: Among the 12 antibiotics included in the analysis (meropenem excluded for absence of resistance), resistance prevalence ranged from 1.1% (colistin) to 59.9% (amoxicillin). Colistin was subsequently also excluded from ML modeling due to insufficient resistant isolates (n = 3), leaving 11 antibiotics for prediction. The best predictive performance was observed for ciprofloxacin (AUROC 0.76 [95% CI 0.74–0.77]; F1 0.54; MCC 0.38) and ceftazidime (AUROC 0.68 [0.65–0.71]; F1 0.36; MCC 0.29), using 13 and 37 RFECV-selected features, respectively. Amoxicillin achieved the highest F1-score (0.76), driven by high recall (0.98) but modest AUROC (0.58). No meaningful predictive signal was detected for amikacin, cefepime, or tigecycline (AUROC ≤ 0.57, F1 ≤ 0.17), attributable to extreme class imbalance, and no robust multi-peak resistance signature was detected in this dataset. McNemar’s test confirmed that RF significantly outperformed Logistic Regression for all antibiotics (p < 0.01), while Gradient Boosting performed comparably to RF for ciprofloxacin (p = 0.17) and ceftazidime (p = 0.28). Frozen extracts (T₂) produced lower spectral similarity and were excluded from model training; the aligned T₁₊₃ dataset yielded the most stable performance across metrics. Conclusions: Machine learning analysis of MALDI-TOF spectra enables reproducible AMR prediction for selected antibiotics in E. coli, with ciprofloxacin and ceftazidime showing the strongest signal. Nested isolate-level cross-validation, multi-model comparison with statistical testing, and open-source code provide a transparent, reproducible foundation for integrating ML-assisted MALDI-TOF analysis into diagnostic AMR surveillance. Extract storage at −20 °C degrades spectral quality and should be avoided in ML training workflows. Full article

The 6 February 2023 Kahramanmaraş earthquakes caused widespread structural damage and highlighted the need for rapid building-level decision support in post-earthquake assessment. This study presents an explainable ensemble learning framework for seismic damage prediction using 16,611 building-level field observations from Kırıkhan, Hatay, Türkiye. The original damage records were reorganized into three operational classes: No-Damage, Slight–Moderate, and Heavy–Collapse. Eight tree-based ensemble models, LightGBM, CatBoost, XGBoost, Random Forest, Extra Trees, Gradient Boosting Machine, AdaBoost, and HistGradientBoosting, were evaluated under a consistent protocol using class-weighting strategies where supported, with Balanced Accuracy as the primary metric. LightGBM and Random Forest achieved the joint-highest Balanced Accuracy value (0.650). Random Forest produced the strongest agreement-based metrics, while LightGBM remained closely competitive and was selected as the representative model for explainability because of its balanced class-wise behavior. CatBoost achieved the highest Heavy–Collapse recall (0.729), XGBoost achieved the highest Macro-AUC (0.821), and GBM produced the highest Overall Accuracy (0.658), showing that model ranking varied by evaluation criterion. SHapley Additive exPlanations identified building age, lithology, number of floors, structural system, plinth area, and proximity to faults and surface ruptures as key contributors. The remaining classification uncertainty, particularly among adjacent damage states, indicates that the framework is best interpreted as a complementary decision-support tool for preliminary screening and prioritization before final safety decisions or official damage assessment. Full article

Parking space availability in densely populated areas has been declining for decades, while extending existing facilities is often not feasible. This paper proposes assigning vehicles to parking spots based on their predicted duration of stay as a means to increase the customer satisfaction of an existing facility. Real-world parking data from an Austrian shopping mall is analyzed to identify vehicle and contextual features that correlate with parking duration, and both regression and classification models are trained to estimate a visitor’s likely stay time upon arrival. A discrete-event simulation then compares four assignment strategies against an unguided baseline, covering a simple nearest-spot approach, a machine learning-based strategy, and a perfect-knowledge upper bound. The primary efficiency gain stems from the elimination of driver search time, which any structured assignment achieves regardless of prediction quality. Duration-aware placement provides a smaller, additional benefit at higher occupancy levels by reserving spaces near the entrance for short-stay visitors, and the gap between the classifier-based strategy and the perfect-knowledge bound remains moderate, confirming that even imperfect predictions yield a meaningful share of the theoretically achievable improvement. It must be noted, however, that the predictive accuracy of the models remains limited: regression errors are near 60 min and classifier accuracy is only modestly above chance, reflecting the fundamental difficulty of inferring individual visitor intent from observable vehicle and arrival features alone. This limitation constrains the practical applicability of duration-aware assignment and should be considered carefully in any real-world deployment decision. Full article

Diabetes mellitus is strongly associated with cardiovascular dysfunction and remains one of the leading contributors to morbidity and mortality worldwide. Early identification of diabetes-related cardiovascular alterations is essential for timely risk stratification and personalized clinical management. In the present study, an interpretable machine learning framework for diabetes classification in patients with cardiovascular disease was developed using routinely available clinical, biochemical, renal, and echocardiographic parameters. A retrospective dataset consisting of 131 cardiovascular patients was included in the final analysis, comprising 65 patients with diabetes mellitus and 66 patients without diabetes. Demographic, metabolic, renal, and cardiovascular variables, including age, body mass index (BMI), glycated hemoglobin (HbA1c), glucose concentration, estimated glomerular filtration rate (eGFR), troponin level, heart rate, and left ventricular ejection fraction (EF), were included in the analysis. Multiple supervised machine learning algorithms, including Logistic Regression, Support Vector Machine (SVM), Gradient Boosting, and Random Forest, were implemented and compared using repeated stratified cross-validation. Among the evaluated models, Random Forest demonstrated the highest classification performance, achieving a mean ROC AUC of 0.880 ± 0.050. Statistical analysis revealed significantly elevated HbA1c, glucose, and troponin levels together with reduced ejection fraction values in diabetic patients. Explainable artificial intelligence analysis using SHAP and partial dependence plots identified glucose concentration, HbA1c, age, and renal function as the dominant contributors to diabetes classification. Nonlinear relationships between metabolic and cardiovascular variables were additionally observed. The obtained findings demonstrate that interpretable machine learning approaches can provide effective discrimination between diabetic and non-diabetic cardiovascular patients while maintaining clinically meaningful interpretability. The proposed framework may contribute to future intelligent clinical decision-support systems and personalized cardiovascular risk assessment strategies. Full article

Carbonation in CO${}_2$-storage solid-waste backfill materials couples CO${}_2$ transport, mineral reaction, and strength evolution, making carbonation-front prediction and transport–reaction inference important for evaluating sequestration performance. This study proposes an evidence-ranked, physics-guided inverse-learning framework for carbonation-front prediction, auxiliary strength reconstruction, PDE-residual assessment, and practical-identifiability analysis. The framework represents carbonation using group-conditioned latent fields of effective CO${}_2$ concentration and remaining reactive capacity, maps latent carbonation degree to measured depth through a differentiable front operator, and reconstructs unconfined compressive strength through a supervised auxiliary head. Empirical front laws and reaction–diffusion physics-informed neural-network variants were evaluated using held-out ranking, repeated stratified splits, residual-weight sweeps, front-operator threshold and smoothing-coefficient sensitivity checks, profile-likelihood and Fisher-information diagnostics, and controlled synthetic tests. Results show that the grouped Weibull front law achieved the best short-range carbonation-depth interpolation, while the retained constant-diffusion PINN was used as a diagnostic formulation within the physics-guided family to improve auxiliary strength reconstruction and to evaluate residual consistency, front-threshold selection, parameter sharing, and inverse-parameter behavior rather than to replace the empirical depth regressor. Increasing the PDE-residual weight substantially reduced residual magnitudes, but profile-likelihood and Fisher-information diagnostics indicated strong parameter trade-offs; the fitted diffusion, reaction, depletion, and diffusion–decay quantities are therefore interpreted as effective, observation-conditional parameters rather than unique material constants. The proposed framework provides a prediction-first and attribution-aware approach for analyzing carbonation evolution in solid-waste backfill materials and supports coordinated assessment of front advancement, strength response, and transport–reaction behavior, while explicitly delimiting the generalization and physical interpretation that can be supported by sparse literature-derived observations. Full article

This study examines whether ESG rating disagreement is a leading indicator of corporate greenwashing and how digital transformation (DTI) moderates this relationship through disclosure and performance channels. Using 8111 firm-year observations from Chinese A-share companies (2012–2022), we employ two-way fixed-effects panel regression complemented by Bayesian-optimised machine learning models interpreted through SHAP. Aggregate rating disagreement is a strong and robust predictor of greenwashing. Channel decomposition reveals asymmetric DTI moderation: the disclosure channel amplifies greenwashing risk as digitally advanced firms expand reporting capacity to widen the gap between disclosed and actual ESG performance (bloom_DTI: β = +0.2471, p < 0.01), while the performance channel attenuates greenwashing risk as digital operational monitoring translates substantive performance into a measurable reduction (hua_DTI: β = −0.2804, p < 0.01). This pattern is robust across ownership structure, pollution intensity, and region. Machine learning analysis confirms the econometric findings and reveals nonlinear threshold effects invisible to panel regression. This asymmetric channel mechanism contributes to the ESG rating divergence literature and has implications for disclosure regulation and ESG-based investment screening. Full article

Accurate retrieval of cloud liquid water content (LWC) and ice water content (IWC) vertical profiles remains limited by strong vertical variability and nonlinear dependencies among observed variables. Ground-based cloud radar reflectivity and microwave radiometer-derived thermodynamic profiles provide complementary constraints, but their joint use requires consistent time–height matching and bias-controlled predictors. This study develops a vertically structured machine-learning framework that explicitly represents profile-level dependencies by constructing vertical-structure-enhanced features to encode local gradients and contextual information, integrating multiple tree-based learners with heterogeneous configurations through a profile-aware stacking strategy, and introducing a profile-level refinement step to suppress layer-to-layer inconsistencies. The framework is evaluated using year-round Cloudnet observations from the Lindenberg site, where IWC RMSE decreases from 0.0152 g m⁻³ to 0.0092 g m⁻³ with R² increasing from 0.412 to 0.784, and LWC RMSE decreases from 0.0786 g m⁻³ to 0.0591 g m⁻³ with R² increasing from 0.303 to 0.606. Additional boundary-region evaluation shows that the improvement is particularly evident near radar-derived cloud boundaries, where cloud structure and hydrometeor content may vary rapidly with height. These results indicate that treating cloud retrieval as a vertically structured learning problem reduces inconsistencies inherent in pointwise models and establishes a data-driven baseline for incorporating vertical constraints into atmospheric profile retrieval. Full article

Safe-separation-and-collision-avoidance unmanned aerial vehicle (UAV) swarms are increasingly used for inspection, emergency response, environmental monitoring, and search-and-rescue support in cluttered airspace where communication links may be delayed, degraded, or intermittently unavailable. These applications require heterogeneous vehicles to maintain situational awareness, allocate tasks, and avoid hazards under partial observability and changing team topology. To address these challenges, this paper proposes a Hierarchical Graph-Attention Multi-Agent Reinforcement Learning architecture (HG-MARL) for safe-separation-and-collision-avoidance heterogeneous UAV swarm coordination. The proposed framework decomposes the task into high-level resource allocation and low-level local-control execution, uses graph attention for changing swarm topology, and applies Transformer memory, action masking, potential-field reward shaping, and domain-randomized simulation training. In the multi-scenario simulation summaries, HG-MARL achieves 92.9%, 89.8%, and 82.6% task success in Scenarios A–C, respectively, improving upon MAPPO by 15.1, 21.4, and 20.1 percentage points. Summary-statistic Welch tests show that all six HG-MARL comparisons against MAPPO and QMIX yield $p<0.01$ with large effect sizes. Fair-control, reward-sensitivity, communication-degradation, safety-ablation, training-stability, latency, and transfer-oriented stress tests further support the contributions of the integrated architecture. The validation scope is simulator-based, with platform-level flight/HIL evaluation discussed as future work. These results suggest that HG-MARL is a promising simulation-validated framework for civilian UAV swarm coordination in collision-and-separation-critical and communication-degraded environments. Full article

The rapid development of the low-altitude economy demands comprehensive electromagnetic spectrum awareness. However, constructing a comprehensive radio environment map (REM) in this scenario is challenging, as spectrum sensing data collected by unmanned aerial vehicles (UAVs) in complex low-altitude environments is typically sparse, fragmented, and non-uniformly distributed across the high-dimensional space of time, frequency, and 3D space. To address these issues, this study proposes a rank-adaptive Bayesian tensor ring completion (Ra-BTRC) framework. The method models the low-altitude electromagnetic environment as a unified five-dimensional (5D) spectrum tensor. It then employs tensor ring (TR) decomposition to capture latent high-order correlations across all dimensions. To overcome the sensitivity of conventional TR methods to predefined ranks, Ra-BTRC introduces sparsity-inducing priors on the TR core factors, enabling variational Bayesian inference to learn observation uncertainty and infer effective TR ranks from sparse measurements without manually fixing the TR rank. Simulations demonstrate that Ra-BTRC significantly outperforms existing TR-based baselines, achieving more than 10 dB MMSE improvement at a 5% sampling rate while accurately recovering local spectrum structures and temporal dynamics. The proposed approach provides a robust and scalable solution for reliable global low-altitude spectrum cognition under stringent sensing budgets. Full article