HeteroM-ILT: hybrid CNN and state space model for high-fidelity mask optimization

Critical infrastructures increasingly rely on AI-generated content (AIGC) for monitoring, decision support, and autonomous control. This dependence creates new attack surfaces: forged maintenance imagery, manipulated diagnostic scans, or spoofed sensor visualisations can trigger unsafe actions, regulatory violations, or systemic disruption. This paper proposes a post-quantum watermarking framework designed for critical infrastructure security. We embed robust provenance markers directly into the latent space of diffusion models, rather than at the pixel level, and reinforce them using error-correcting codes (ECC) to ensure watermark recoverability even after aggressive distortions such as compression, cropping, noise injection, and filtering. To secure watermark keys in transit and at rest, we integrate Kyber, a lattice-based key encapsulation mechanism standardised for post-quantum cryptography, to protect the watermark stream key against quantum-enabled interception. The resulting scheme (i) preserves visual fidelity, (ii) supports reliable forensic attribution and auditability under hostile conditions, and (iii) remains cryptographically secure in the post-quantum era. Experiments show that the proposed ECC-hardened latent watermarking achieves consistently high extraction accuracy across diverse attacks while maintaining image quality, outperforming state-of-the-art diffusion watermarking baselines. We position this watermarking–encryption pipeline as an enabling mechanism for privacy-aware traceability, zero-trust validation, and quantum-resilient content governance in next-generation critical infrastructure.

Mining under the Ordovician limestone aquifer is severely threatened by coal seam floor water inrush, which endangers mining safety. Traditional fixed-weight evaluation models for coal floor water inrush risk have prominent limitations: they ignore the internal variability among index values, easily distorting prediction results and failing to meet the demand for accurate risk assessment. To address this issue, this study proposes a dynamic variable weight evaluation model for coal floor water inrush risk based on the Improved Fuzzy Analytic Hierarchy Process (IFAHP) and Entropy Weight Method (EWM). First, after comprehensive analysis of hydrogeological data, six key evaluation indicators were selected, including Ordovician limestone water pressure, Ordovician limestone water abundance, aquifuge thickness, fragile rock ratio, fault fractal dimension, and floor failure depth. Subjective weights from IFAHP and objective weights from EWM were then combined to obtain comprehensive weights. Subsequently, the K-means algorithm was applied for unified clustering analysis of the indices to define variable weight intervals and construct the dynamic variable weight model. Based on this model, ArcGIS was used to classify Ordovician limestone water inrush risk into four levels and generate a risk zoning map for the 16<sup>th</sup> coal seam floor. Validation with known water inrush points and comparative analysis with the constant-weight model demonstrated that the proposed IFAHP-EWM dynamic variable weight method has higher prediction accuracy and better spatial adaptability. Additionally, 3D visualization of the risk zoning map was developed to enhance practical application. This model provides a reliable theoretical basis and practical tool for improving the safety of coal seam floor mining under Ordovician limestone aquifers.

The proliferation of unmanned aerial vehicles (UAVs) poses escalating security threats across critical infrastructures, necessitating robust real-time detection systems. Existing vision-based methods predominantly rely on single-modality data and exhibit significant performance degradation under challenging scenarios. To address these limitations, we propose DCAM-DETR, a novel multimodal detection framework that fuses RGB and thermal infrared modalities through an enhanced RT-DETR architecture integrated with state space models. Our approach introduces four innovations: (1) a MobileMamba backbone leveraging selective state space models for efficient long-range dependency modeling with linear complexity O(n); (2) Cross-Dimensional Attention (CDA) and Cross-Path Attention (CPA) modules capturing intermodal correlations across spatial and channel dimensions; (3) an Adaptive Feature Fusion Module (AFFM) dynamically calibrating multimodal feature contributions; and (4) a Dual-Attention Decoupling Module (DADM) enhancing detection head discrimination for small targets. Experiments on Anti-UAV300 demonstrate state-of-the-art performance with 94.7% mAP@0.5 and 78.3% mAP@0.5:0.95 at 42 FPS. Extended evaluations on FLIR-ADAS and KAIST datasets validate the generalization capacity across diverse scenarios.

Dense temporal measurements of physiological health, using simple and consistent assays, are essential to characterize biological processes associated with aging and evaluate the effectiveness of interventions on these processes. We measured body weight in 960 genetically diverse female mice, every 7-10 days over the full course of their lifespan. We used a state space model to characterize the trajectories of body weight throughout life and derived novel traits capturing the dynamics of body weight, 10 of which were both heritable and associated with lifespan. Genetic mapping of these body weight-derived traits identified 5 genomic loci, none of which were previously mapped to body weight. We observed that the ability to maintain stable body weight, despite fluctuations in energy intake and expenditure, was positively associated with lifespan in an age-dependent manner and mapped to a genomic locus linked to energy homeostasis. Our results highlight how dense longitudinal measurements of physiological phenotypes offer new insights into the biology of aging.

Abstract A precise methodology for determining the growth mode of oxide layers on metallic materials at high temperatures is proposed. The approach combines sequential isotopic oxidation tests (using 16 O and 18 O isotopes) with secondary ion mass spectrometry (SIMS and nanoSIMS) analyses. NanoSIMS provides high-resolution localisation of oxygen diffusion pathways and oxide growth zones. However, its limited accessibility and specialised instrumentation can pose practical constraints. In contrast, dynamic SIMS offers broader accessibility and the ability to directly quantify oxygen isotope ratios across depth profiles. The detection of both conventional atomic (O − ) and diatomic (O 2 − ) oxygen signals in dynamic SIMS analysis proved highly effective in offering insights on oxide growth mode, closely replicating nanoSIMS results. The diatomic signal analysis complements the atomic signal data by improving the understanding of oxidant transport within the oxide layer. The methodology was validated through its application to a Co-10Cr alloy oxidised at 900 °C in O 2 , under sequential exposures to 16 O and 18 O isotopes. Both SIMS and nanoSIMS revealed the formation of a duplex oxide layer, consisting of an outer layer formed by outward Co cation diffusion and an inner layer growing by inward oxygen penetration, particularly in the grain-boundary regions of the outer oxide layer. The alloy is proposed to oxidise according to the Available Space Model.

Many real-world problems feature nonlinear dynamic processes. Classical mathematical models may be adequate to describe a single dynamic process in isolation, but can be easily undermined by two natural and simple kinds of phenomenological variations: the emergence (or activation) of an additional dynamic process, and events that affect the parameters of an active process. COVID-19 data offers an important case study expressing these phenomenological variations that deeply challenge the classical SIR epidemiological model, and call for novel mathematical methods to detect and adapt to these critical variations. We address the modeling issues with a novel mathematical framework that reenvisions data as a mixture of multiple causal generating processes, each subject to possible parameter change-points. The new viewpoint extends nonlinear classical models in a manner that overcomes many of these types of phenomenological variations and enables a highly adaptive modeling closely linked to causal events. The new model space unifies a wider class of dynamics and is particularly effective at fitting multi-surge data and explaining key causal events related to surge origination. To demonstrate, we construct a mixture of logistic models termed the Adaptive Logistic Model (ALM), and then formulate appropriate nonlinear least squares optimization and regularization goals, and then apply ALMto data. To validate the approach, we return to COVID-19 forecasting (for case count), and compare ALM directly to other forecasting methods. ALM forecast accuracy is competitive with all leading forecast methods, but its greatest utility may be in how it detects changing dynamics (change-points) and retains far fewer but more interpretable parameters relating naturally to cause and intervening change. The method can be applied more generally as it adapts well to the multi-generative nature of many time series data problems. We demonstrate ALM robustness through data experiments in hydrology, economics, cybersecurity, and social media.

The Prognostics and Health Management (PHM) field faces significant challenges due to fragmented benchmarks, inconsistent evaluation protocols, and limited accessibility to comprehensive frameworks, particularly in the era of large-scale data and foundation models. To address these critical limitations, we introduce PHM-Vibench, a unified, extensible, and modular benchmarking platform for vibration-based PHM research. PHM-Vibench features a novel architecture that decouples the PHM pipeline into distinct data, model, task, and trainer factories, enabling flexible instantiation and customization of specific PHM workflows. The platform integrates comprehensive 20+ datasets with standardized protocols. It supports diverse PHM tasks including fault diagnosis, remaining useful life prediction, and anomaly detection. The framework addresses complex scenarios such as domain generalization, cross-system transfer, few-shot learning. Grounded in the Unified PHM Problem (UPHMP) framework with seven fundamental spaces: domain knowledge space (P), data space (D), task space (T), model space (M), loss function space (L), protocol space (Π), and evaluation metric space (E), PHM-Vibench enables systematic problem formalization and reproducible experimentation. The platform accommodates both traditional machine learning models and foundation models, with extensive experimental validation demonstrating superior cross-domain performance. PHMVibench addresses the standardization challenges in PHM research and provides a comprehensive solution for benchmarking and advancing the field. The platform is openly available at https://github.com/PHMbench/PHM-Vibench.

Advancements in artificial intelligence have significantly enhanced communication for individuals with hearing impairments. This study presents a robust cross-lingual Sign Language Recognition (SLR) framework for Turkish, American English, and Arabic sign languages. The system utilizes the lightweight MediaPipe library for efficient hand landmark extraction, ensuring stable and consistent feature representation across diverse linguistic contexts. Datasets were meticulously constructed from nine public-domain sources (four Arabic, three American, and two Turkish). The final training data comprises curated image datasets, with frames for each language carefully selected from varying angles and distances to ensure high diversity. A comprehensive comparative evaluation was conducted across three state-of-the-art deep learning architectures—ConvNeXt (CNN-based), Swin Transformer (ViT-based), and Vision Mamba (SSM-based)—all applied to identical feature sets. The evaluation demonstrates the superior performance of contemporary vision Transformers and state space models in capturing subtle spatial cues across diverse sign languages. Our approach provides a comparative analysis of model generalization capabilities across three distinct sign languages, offering valuable insights for model selection in pose-based SLR systems.

Timely and accurate detection of forest fires through unmanned aerial vehicle (UAV) remote sensing target detection technology is of paramount importance. However, multiscale targets and complex environmental interference in UAV remote sensing images pose significant challenges during detection tasks. To address these obstacles, this paper presents FF-Mamba-YOLO, a novel framework based on the principles of Mamba and YOLO (You Only Look Once) that leverages innovative modules and architectures to overcome these limitations. Specifically, we introduce MFEBlock and MFFBlock based on state space models (SSMs) in the backbone and neck parts of the network, respectively, enabling the model to effectively capture global dependencies. Second, we construct CFEBlock, a module that performs feature enhancement before SSM processing, improving local feature processing capabilities. Furthermore, we propose MGBlock, which adopts a dynamic gating mechanism, enhancing the model’s adaptive processing capabilities and robustness. Finally, we enhance the structure of Path Aggregation Feature Pyramid Network (PAFPN) to improve feature fusion quality and introduce DySample to enhance image resolution without significantly increasing computational costs. Experimental results on our self-constructed forest fire image dataset demonstrate that the model achieves 67.4% mAP@50, 36.3% mAP@50:95, and 64.8% precision, outperforming previous state-of-the-art methods. These results highlight the potential of FF-Mamba-YOLO in forest fire monitoring.

ABSTRACT The Test of Chinese as a Heritage Language, or Huawen Shuiping Ceshi (HC), is a newly developed proficiency test by a university in Guangzhou, China. It defines “Heritage Language” as a family-transmitted language that is non-dominant in broader society and often incompletely acquired. The reading test is one of its three subtests. To investigate the cognitive patterns of learners of Chinese as a heritage language and to provide diagnostic assessment for test takers and educational institutions, the Rule Space Model (RSM) was implemented to conduct a diagnostic research of 236 test takers’ responses on the reading test (Level 3). The research results indicate that: (1) the reading attributes of Chinese as a Heritage Language (Reading) were relatively accurately determined, and the hierarchical relationships of attributes were supported by the empirical data; (2) there were 12 mastery patterns of reading attributes for learners of Chinese as a heritage language. The RSM was successfully implemented in the diagnostic assessment of the reading test. This study can enhance the understanding of Chinese reading ability, advance the application of cognitive diagnosis in Chinese reading tests, and offer more detailed insights for reading instruction.

ABSTRACT For the analysis of nonlinear or non‐Gaussian state space models (SSMs), extended Kalman filters and particle filters (PFs) are proposed in the literature. Although these filters allow to formulate SSMs that are much more flexible compared to the linear Gaussian model, they are still based on parametric distributions. In this paper, a novel PF is proposed for the analysis of high‐frequency time series with heavy tails and outliers such as GPS data, road sensor data, climate data, social media data, and data on stock prices. A neural network (NN) is trained using multiple time series to obtain a nonparametric approximation of the probability density function for the observation equation of the SSM and is combined with a PF to obtain estimates for the unobserved states of a local level model. This results in the so‐called neural network particle filter (NNPF). We illustrate the accuracy gains from our proposed method in an extended simulation where time series are generated under the assumption of a local level model and Gaussian, Student's , noncentral Student's , and Poisson distributions for the observation equation. The proposed NNPF outperforms existing filters, particularly in the case of continuous distributions with heavy tails and outliers. The proposed NNPF is applied to a real‐life application using vehicle minute counts obtained with road sensors.

A causal bidirectional selective state space model for imaging genetics in neurodegenerative diseases.

State space models have recently emerged as a powerful framework for long sequence processing, outperforming traditional methods on diverse benchmarks. Fundamentally, state space models can generalize both recurrent and convolutional networks and have been shown to even capture key functions of biological systems. Here we report an approach to implement SSMs in energy-efficient compute-in-memory hardware to achieve real-time, event-driven processing. Our work re-parameterizes the model to function with real-valued coefficients and shared decay constants, reducing the complexity of model mapping onto practical hardware systems. By leveraging device dynamics and diagonalized state transition parameters, the state evolution can be natively implemented in crossbar-based compute-in-memory systems combined with memristors exhibiting short-term memory effects. Through this algorithm and hardware co-design, we show the proposed system offers both high accuracy and high energy efficiency while supporting fully asynchronous processing for event-based vision and audio tasks.

Optimizing indoor lighting in high-ceiling spaces requires balancing functionality, comfort, and energy efficiency across multiple indicators. Lighting rules under different strategies were explored via quantitative analysis and global sensitivity analysis, and a multi-criteria decision-making model was built using principal component analysis (PCA). Results indicate that ceiling downlights have the greatest overall impact on illuminance uniformity and unified glare rating, offering the largest optimization potential. Low-level accent lighting and indirect lighting can reduce installed power while maintaining high efficacy through zoned design of critical areas. The resulting sensitivity matrix linking each index to each strategy supports early-stage scheme selection aligned with project priorities. The PCA-based decision model integrates correlated indices into a single score, enabling objective comparison of alternatives and identification of optimal solutions for large, high-ceiling interiors. It therefore helps designers trade off uniformity, glare control, and power density, and provides a transparent basis for choosing a strategy early.

Most of the intracyclic variability in the large-scale solar magnetic field comes from the equatorial dipole component of the solar magnetic field. The equatorial dipole component is highly sensitive to the longitude distribution of the active regions. We quantify the effect of individual active regions on the large-scale solar magnetic field of the solar cycle 24. We study the effect of the longitude distribution of active regions on the strength of the large-scale dipole component. We used a surface flux transport (SFT) model to simulate the evolution of individual active regions and quantified their effect on the large-scale magnetic field using the recently developed vector sum method. We took advantage of the longitudinal translational invariance of the SFT model and compared the observed solar cycle 24 to the 10 000 simulations of the solar cycle 24 using randomized longitudinal source locations, but otherwise identical flux emergence. We find that taking into account both the axial and equatorial components of the vector sum characterizing the global solar magnetic field sets better constraints on the parameter space of the SFT model than, for example, using the axial dipole moment alone as an optimization metric. We studied the maximum of cycle 24 and identified the recurrent and localized flux emergence in the southern hemisphere as the main culprit behind the rapid strengthening of the large-scale magnetic field in late 2014. We find that during the declining phase of the solar cycle, the strength of the large-scale magnetic field stayed above the median level of randomized simulations (p $<$ 0.027) for 42 subsequent rotations (from September 2014 to November 2017). This indicates that the longitudinal distribution of active regions is not random and, rather, that it demonstrates a tendency for some regions to emerge at longitudes where their equatorial components reinforce the large-scale equatorial field.

The design of selective kinase inhibitors remains a formidable challenge due to the high structural conservation of the ATP-binding site across the kinome. While modern generative AI has enabled rapid exploration of chemical space, many advanced models operate as black boxes, obscuring the chemical rationale behind design choices and limiting interpretability. To explore these bottlenecks, we present a modular, generative framework for de novo design of SRC kinase inhibitors, integrating ChemVAE-based latent space modeling, a chemically interpretable Kinase Inhibition Likelihood (KIL) scoring function, Bayesian optimization, and cluster-guided local neighborhood sampling. The results demonstrate that kinase inhibitors spontaneously organize into a coherent, low-dimensional manifold in latent space, with SRC acting as a structural "hub" that enables rational scaffold transformation. Our local neighborhood sampling-based approach successfully converts inhibitors from other kinase families (notably LCK) into novel SRC-like chemotypes, with LCK-derived molecules accounting for ∼40% of high-similarity outputs. Critically, we expose a fundamental representation gap: despite aromatic ring count being a top KIL feature, SMILES-based generation systematically fails to access multi-ring pharmacophores characteristic of clinical kinase inhibitors. This limitation cannot be overcome by scoring refinement alone, demanding topology-aware representations. Our framework also demonstrates that unbiased exploration paired with cluster-guided sampling outperforms active-biased optimization, which traps search in narrow local optima. By exposing representational gaps and showcasing scaffold-aware navigation of latent space, this study argues for hybrid systems that combine the diagnostic transparency of interpretable machine learning frameworks with the generative power of modern architectures.

Does tourism reduce socio-economic inequality? Insights from the pandemic based on space and time models for the EU

Building extraction underpins land-use assessment, urban planning, and disaster mitigation, yet dense urban scenes still cause missed small objects, target adhesion, and ragged contours. We present High-Resolution-Mamba (HR-Mamba), a high-resolution semantic segmentation network that augments a High-Resolution Network (HRNet) parallel backbone with edge-aware and sequence-state modeling. A Canny-enhanced, median-filtered stem stabilizes boundaries under noise; Involution-based residual blocks capture position-specific local geometry; and a Mamba-based State Space Models (Mamba-SSM) global branch captures cross-scale long-range dependencies with linear complexity. Training uses a composite loss of binary cross entropy (BCE), Dice loss, and Boundary loss, with weights selected by joint grid search. We further design a feature-driven adaptive post-processing pipeline that includes geometric feature analysis, multi-strategy simplification, multi-directional regularization, and topological consistency verification to produce regular, smooth, engineering-ready building outlines. On dense urban imagery, HR-Mamba improves F1-score from 80.95% to 83.93%, an absolute increase of 2.98% relative to HRNet. We conclude that HR-Mamba jointly enhances detail fidelity and global consistency and offers a generalizable route for high-resolution building extraction in remote sensing.

Denmark’s aging population is increasingly affected by loneliness, a condition with severe public health and economic consequences, including an estimated annual cost of 2.2 billion DKK and approximately 770 excess deaths. Cohousing communities offer a promising approach to counteract these effects by fostering social networks that enhance health and well-being among older adults. Yet, architects often lack an evidence-based understanding of how spatial design supports social interaction. This study conducts a comparative analysis of four Danish cohousing developments to identify spatial features that promote social connectedness. The findings inform the development of the Socially Enhancing Environmental Spaces (SEES) Model, which provides architects and policymakers with a framework for designing built environments that encourage social engagement and improve quality of life in later life.