Search Results (454)

We develop a semi-parametric framework for representing discrete probability mass functions through orthogonal polynomial representations. Classical count models, such as the Poisson and negative binomial distributions, impose restrictive structural assumptions that often fail to accommodate empirical features including heavy overdispersion, multimodality, and nonstandard tail behavior. To address these limitations, we introduce a linear-tilt model constructed from orthonormal polynomial systems associated with Poisson and negative binomial baselines, namely the Charlier and Meixner families. The proposed representation improves the baseline distribution using additional information from empirical moments. This allows the distribution to flexibly adjust its shape, capturing differences in skewness and kurtosis. We establish theoretical properties of the expansion within a weighted Hilbert space formulation, where the coefficients arise as orthogonal projections that can be expressed as expectations of the corresponding polynomial basis functions. In addition, we analyze approximation behavior and provide numerical bounds on the resulting numerical error and convergence properties of truncated approximations. The practical relevance of the proposed methodology is illustrated through applications to several empirical datasets, demonstrating its ability to capture complex distributional structures while preserving a tractable semi-parametric form. Full article

Visually rich document understanding (VrDU) models rely on tightly coupled textual, layout, and visual representations. In real-world deployments, these models must continuously adapt to new document domains over time. However, naïve sequential fine-tuning leads to severe catastrophic forgetting due to shared parameters and strong cross-task interference. Existing continual learning approaches either constrain parameter updates, preserve output distributions, or uniformly suppress gradient directions associated with previous tasks. While effective in limited settings, these strategies fail to balance stability and plasticity in large multimodal transformers. We propose AMD-Proj, an adaptive memory-driven selective gradient projection framework for continual learning in document understanding. It models task knowledge using specific gradient subspaces and adaptively modulates incoming gradients based on their alignment with this memory, selectively blocking interfering directions while reinforcing reusable ones. An efficient truncated SVD mechanism with online subspace merging ensures bounded memory usage and scalability to large transformer-based architectures. We evaluate AMD-Proj on four VrDU benchmarks (FUNSD, SROIE, CORD, and BuDDIE) under a task-incremental learning setting using LayoutLMv2 and LayoutLMv3 backbones. Results show that AMD-Proj reduces catastrophic forgetting and improves F1-based stability over EWC, GPM, LwF, OWM, CUBER, TRGP and parameter-efficient fine-tuning methods. Extensive mechanistic analyses, including gradient spectrum decomposition and layer-wise reuse versus block dynamics, provide insight into how selective gradient projection controls optimization geometry during continual adaptation. These findings establish selective gradient projection as a principled and interpretable approach for continual learning in visually rich document understanding. Full article

Traditional wood identification models are vulnerable to out-of-distribution (OOD) substitution in the global timber trade. In response to this issue, this study presents a dual-stage cascade authentication architecture using in situ Raman spectroscopy and machine learning. First, a physically informed preprocessing strategy, integrating adaptive truncation (>1749 cm⁻¹) and first-derivative filtering, is developed to extract a 1309-dimensional composite feature matrix. This step effectively decouples non-linear fluorescence and converts physical detector saturation into highly discriminative features. To mitigate data leakage, the system utilizes a cross-validated Random Forest engine for Stage-1 closed-set discriminative screening. Subsequently, it cascades a high-dimensional One-Class Support Vector Machine (OCSVM) for Stage-2 open-set non-linear boundary verification in the Reproducing Kernel Hilbert Space. This design avoids the “variance trap” of traditional linear dimensionality reduction (e.g., PCA), preserving weak but critical secondary metabolite signals. Under a controlled OOD benchmarking scenario involving three taxonomically and chemically similar substitute species, the optimized Stage-1 engine maintains a 91.67% closed-set accuracy on known species. Crucially, Stage-2 verification achieves an open-set detection AUROC of 0.9722 and limits the FPR95 to 3.33%. Feature importance mapping indicates that the model effectively incorporates macroscopicoptical surrogate features (e.g., fluorescence decay boundaries) for decision-making. Overall, this study offers a robust, controlled non-destructive approach for real-world wood authenticity verification. Full article

This study evaluates the mechanical performance of FDM-printed poly(lactic acid) (PLA) structures joined using a portable Friction Stir Welding (FSW) device. A non-destructive optical band method was employed to assess weld homogeneity and material flow consistency. The influence of substrate infill density (15% and 100%) and tool pin geometry (cylindrical and truncated conical) was systematically analyzed. Results indicate that substrate density is the primary determinant of joint integrity; 100% infill specimens demonstrated superior structural homogeneity and consistent intensity profiles, whereas 15% infill specimens exhibited significant intensity fluctuations and poor consolidation, even with the addition of filler material. The mechanical evaluation revealed that the use of a tool pin is essential for effective load transfer, as specimens welded without internal agitation achieved only baseline tensile strengths of approximately 4 MPa. Among the pin-driven configurations, the cylindrical geometry outperformed the truncated conical design, reaching a peak tensile stress of 8.02 ± 1.42 MPa, corresponding to a joint efficiency of 27% relative to the 100% infill base material, compared to 6.25 ± 1.43 MPa. This performance gap is attributed to the cylindrical pin’s ability to maintain higher shear rates and more uniform pressure distribution at the weld root. These findings demonstrate the feasibility of portable FSW for structural joining of additively manufactured polymers and establish critical processing parameters for the optimization of portable FSW in engineering applications. Full article

We propose a sequential design method aiming at the estimation of an extreme quantile based on a sample of binary data corresponding to peaks over a given threshold. This study is motivated by an industrial challenge in material reliability and consists of estimating a failure quantile from trials whose outcomes are reduced to indicators of whether the specimen has failed at the tested stress levels. The proposed approach relies on a splitting strategy that decomposes the target extreme probability into a product of higher-order conditional probabilities, enabling a progressive exploration of the tail of the distribution through sampling under truncated laws. We consider GEV and Weibull models for the underlying distribution, and the sequential estimation of their parameters is carried out using an enhanced maximum likelihood procedure specifically adapted to binary data, addressing the substantial uncertainty inherent to such limited information. Full article

We consider the problem of Bayesian parameter inference in the Merton structural credit risk model, where the posterior is induced by a jump-diffusion likelihood and the marginal evidence is not available in closed form. To approximate this posterior, we construct a variational family based on triangular sum-of-squares (SOS) polynomial flows, in which each component map is monotone by construction: its diagonal derivative is a positive definite quadratic form on a monomial basis, yielding a closed-form log-Jacobian and explicit gradients with respect to all flow parameters. The symmetric positive definite matrices parametrizing the flow are optimized by intrinsic Riemannian gradient ascent on the positive definite cone equipped with the affine-invariant metric, which preserves feasibility at every iterate without projection. We show that the rank-one Jacobian gradients produced by the SOS structure have unit norm in the affine-invariant metric, establishing a direct algebraic coupling between the transport family and the optimization geometry and implying a universal 1-Lipschitz bound for the log-Jacobian along geodesics. On the likelihood side, we derive exact score identities for all five structural parameters of the Merton model—drift, volatility, jump intensity, jump mean, and jump volatility—through both the Poisson log-normal mixture and the Fourier inversion representations. Strictly positive parameters are handled via exponential reparametrization, and the resulting gradients propagate end-to-end through the flow. We establish uniform truncation bounds on compact parameter sets for the infinite mixture and its associated score series, providing rigorous control over the finite approximations used in practice. The base distribution is chosen to be uniform on ${[0,1]}^5$, whose bounded support ensures uniform control of the monomial basis and stabilizes the polynomial calculus. These ingredients are assembled into a fully explicit modified ELBO with implementable gradients, combining Euclidean updates for vector parameters and intrinsic manifold updates for matrix parameters. Full article

Photonic jets (PJs) generated from mesoscale dielectric particles can achieve sub-diffraction-scale light field constraints and significant near-field intensity enhancement, which have important application value in the fields of nanoimaging, optical sensing, and laser processing. Recent studies show that the axial-extension and transverse-focus characteristics of PJs can be effectively regulated through interface engineering methods, such as using double-layer structures and truncated geometries. Such structures can be referred to as Janus microstructures separated by surface refracted interfaces. However, systematic research on the effect of incident light polarization on the formation and regulation of PJs on the surface interfaces of Janus systems is lacking. In this study, the PJ characteristics under polarization regulation in curved-interface Janus microcylinders are systematically investigated by performing full-wave numerical simulations. The results show that polarization modulation introduces a new degree of freedom for regulating the energy flow distribution and morphology of PJs. An appropriate polarization state can be selected to effectively regulate key characteristic parameters, such as the length, peak intensity, and full width at half maximum of the nanojet, without changing the particle geometry or material composition. This study reveals the synergy between the surface-interface Janus structures and polarization engineering, providing a new physical method for the flexible regulation of PJs in near-field optics. Full article

The Johnson–Mehl–Avrami–Kolmogorov (JMAK) formalism provides a classical framework for describing polymer crystallization kinetics; its applicability under finite-domain confinement requires quantitative assessment. In this work, the influence of one-dimensional geometric restriction on cylindrical growth in polymer thin films is investigated using a stochastic Monte Carlo approach. The model considers site-saturated nucleation on randomly distributed cylindrical nanofibers with constant radial growth velocity under hard-wall boundary conditions. Crystallization kinetics were evaluated through automated segmented regression of the double-logarithmic JMAK representation. Under confinement, the Avrami plot departs from single-slope linearity and exhibits two successive quasi-linear regimes characterized by effective parameter pairs $\left(n_1,\ln k_1\right)$ and $\left(n_2,\ln k_2\right)$. The primary exponent $n_1$ remains thickness-independent, consistent with early-stage radial expansion prior to boundary interaction. The secondary exponent $n_2$ displays a non-monotonic dependence on reduced film thickness, reflecting the competing influence of wall-induced truncation and inter-domain impingement on late-stage transformation. These results support a geometric interpretation in which finite-domain constraints modify the apparent Avrami response through the competing effects of wall-induced truncation and inter-domain impingement and provide a reproducible framework for analyzing dual-regime Avrami behavior in confined crystallization systems. Full article

Rhabdophis lateralis is a snake species within the family Natricidae, which is widely distributed across mainland China, Russia, and Korea. Although this species was once thought to be non-venomous, there are quite a few cases demonstrating its bite could be fatal. In this study, we performed de novo assembly and analysis of the transcriptome data from the Duvernoy’s gland of R. lateralis, aiming to characterize its venom transcriptome and reveal the molecular basis of its toxicity. Among 6196 annotated transcripts, 77 were identified as potential toxin transcripts belonging to 26 toxin families. The most highly expressed toxin family was the SVMP family, accounting for 51.10% of the total toxin expression. The other notable toxins included cysteine-rich secretory proteins (CRISPs, 22.36%), c-type lectins (CTLs and snaclecs, 12.13%), and three-finger toxins (3Ftxs, 6.36%). Phylogenetic analyses indicated that SVMPs, CRISPs, and three-finger toxins (3FTxs) are evolutionarily conserved within Colubridae, whereas CTLs likely arose through convergent evolution. All identified SVMPs were classified as P-III type, with one sequence displaying a unique deletion distinct from conventional truncation patterns. The predominantly expressed CTLs are more likely to combine into dimers, exerting coagulation activity. This study provides an insight into the toxin gene expression in the Duvernoy’s gland of R. lateralis, which will benefit future research into the ecological and pharmacological significance of toxins in the genus Rhabdophis. Full article

Sensing mutual information (SMI) is widely adopted as a performance metric for integrated sensing and communication (ISAC) to enhance both sensing and communication capabilities. However, conventional approaches derive SMI from amplitude and phase, whereas an explicit evaluation of range mutual information (RMI) remains absent. In this paper, we investigate a novel closed-form approximation of RMI for ISAC. We first derive an explicit expression for the posterior probability density function (PDF) of the target range, which is formulated as a function of the signal’s autocorrelation and cross-correlation. Furthermore, we show that under high signal-to-noise ratio (SNR), the estimated range PDF approximates a Gaussian distribution in the sensing-unconstrained scenario and a truncated Gaussian distribution in the sensing-constrained scenario. Finally, we derive closed-form approximations of the RMI in both scenarios under high SNR. In the sensing-unconstrained scenario, the RMI is proportional to the delay interval, root-mean-square bandwidth, and SNR. In the constrained scenario, we obtain a closed-form RMI approximation by introducing an entropy correction term that quantifies the impact of boundary constraints. Additionally, we employ a maximum likelihood estimation (MLE) method to assess range estimation performance. Simulation results validate the accuracy of the theoretical results and the effectiveness of the proposed approximations. Full article

In the insurance literature, accurately predicting extreme losses has been a persistent and important problem. Recently, under the modelling framework of weighted distributions, several finite-mixture size-biased distributions, including size-biased Weibull and size-biased truncated log-normal distributions, have gained popularity for modelling heavy-tailed insurance claim data. In this study, unlike existing models, we explicitly account for the individual heterogeneity commonly observed in insurance claims by treating the order of size-biased weighting as a continuous latent variable, thereby constructing a mixed size-biased distribution. In particular, we study the various distributional properties of the mixed log-normal distribution with a truncated normal prior, which serves as a conjugate prior for the size-biased log-normal model. For applications in non-life insurance, we discuss the Bayesian credibility premium and present an estimation of a regression model via the EM algorithm. We further conduct a real-data analysis using insurance loss data, comparing goodness-of-fit and tail risk measures with those of standard heavy-tailed distributions. Full article

We propose Distribution-Preserving Latent Steganography via Conditional Optimal Transport (DPL-COT), a coverless image steganography framework for latent diffusion models. Unlike classical cover-modifying schemes, DPL-COT embeds a bitstream directly into the initialization noise latent ${\mathbf{z}}_T\sim \mathcal{N}(\mathbf{0},\mathbf{I})$ without model retraining. Our primary objective is high recoverability and a low bit error rate (BER) under deterministic inversion, which is inherently imperfect due to numerical discretization and VAE nonlinearity. To maximize decoding stability, we restrict embedding to the natural tails of the latent prior by selecting the largest-magnitude coordinates, thereby increasing the sign decision margin against inversion drift. To preserve distributional stealth, per-bit target values are analytically derived from truncated Gaussians matching the marginal distribution of the selected coordinates. Conditional 1D optimal transport is applied independently for each bit class, mapping every coordinate to its target value while preserving rank order. We generate 5000 stego images using a pretrained diffusion model and demonstrate a favorable capacity–reliability trade-off (e.g., 4916 bits/image with 0.473% mean BER) and strong robustness to JPEG compression (sub-1% mean BER at $Q=60$). Compared with LDStega, a recent LDM-based scheme reporting 99.28% clean-channel accuracy, DPL-COT achieves 99.53% at a comparable operating point and sustains above-99% accuracy under all tested JPEG quality factors. Latent-space tests further confirm negligible cover–stego distribution shift (mean ${\mathrm{KS}}_2<0.003$, mean $W_1<0.003$), a property not formally addressed by prior methods. Full article

Smart energy networks, including electricity distribution and district heating, are increasingly operated as sensor-enabled infrastructures where maintenance decisions must be made under heterogeneous and time-varying operating conditions. In these settings, time-to-event data are rarely complete; preventive actions and limited observation horizons routinely introduce censoring and truncation, so models and validation procedures must account for partially observed lifetimes to avoid biased inference and misleading performance estimates. This review surveys survival models for predictive maintenance (PdM) and remaining useful life (RUL) estimation, spanning non-parametric, semi-parametric, parametric, and learning-based approaches, with emphasis on censoring-aware formulations and the use of static and time-varying covariates derived from sensor, inspection, and contextual information. A structured taxonomy and a systematic mapping of model families to data types, core assumptions (proportional hazards versus parametric distributional structure), and decision-oriented outputs such as risk ranking, horizon failure probabilities, and RUL distributions are presented. Evaluation practice is also synthesized by covering discrimination metrics, censoring-aware RUL accuracy measures, and probabilistic assessment via proper scoring rules, including the time-dependent Brier score and Integrated Brier Score (IBS). The review provides researchers and practitioners with a practical guide to selecting, fitting, and evaluating survival models for risk-informed maintenance planning in smart energy networks. Full article

Afrotropical inland waters remain poorly studied for rotifer diversity. Here, we provide new distribution data from Botswana and connect these local patterns to continental-scale biogeography using an Africa–Europe occurrence dataset. In Botswana, we analyzed rotifer species richness, functional traits, and environmental drivers using 37 samples from 15 water bodies spanning natural and anthropogenic habitats. We recorded 107 rotifer taxa: 92 identified to species or subspecies level, 14 to genus, and one group of unidentified bdelloids. Seventy taxa (~65%) are new records for Botswana, and one species, Donneria sudzukii, is reported for the first time in Africa. Physicochemical gradients explained community structure, with the first two constrained RDA axes accounting for 40.7% and 23.7% of variation. Axis 1 captured a mineralization gradient linked to total dissolved solids and temperature, whereas Axis 2 reflected oxygen concentration and pH. Traits tracked these gradients: warmer, more mineralized waters were associated with specific trophi types, compact body shapes, and intermediate body sizes, whereas less mineralized, better oxygenated sites were related to smaller taxa and alternative feeding morphologies. To place these trait–environment relationships in a broader geographic context, we then analyzed an Africa–Europe dataset (67,170 records) to quantify latitudinal patterns in thermal classes and morphological traits (geometric body shape and trophi type). Diversity showed clear latitudinal structuring: warm-water genera clustered at low latitudes, only Kellicottia and Didymodactylos had mean distributions above 50° N, and bdelloid families were associated with higher latitudes. Morphological traits also varied with latitude, with trilateral truncated pyramid body shapes and malleoramate trophi occurring closest to the equator. Overall, by combining new species-level data from Botswana with continent-scale occurrence patterns, we link local community assembly to macroecological structure in rotifer functional and biogeographical organization. Full article

Functional time series (FTS) modeling has emerged as a powerful framework for capturing complex temporal dependencies using the functional autoregressive models FAR(p, m) and FARX(p, m, τ). These functional models characterize the evolution of functional observations by incorporating ‘p’ lagged functional responses, ‘m’ truncated dimensions from functional principal component analysis (FPCA), and τ number of scalar covariates with optimal parameter selection guided by the minimization of the functional final prediction error fFPE(p, m). The aim of this study is to propose a computationally efficient FAR model that can integrate a number of functional covariates to achieve a high predictive accuracy in terms of standard out-of-sample accuracy measures. To this end, an integrated functional autoregressive model FAR$\mathbf{X}(p,m,\underline{\mathbf{g}},\tau )$ is developed, where X denotes the exogenous information, this being a lagged or modeled functional profile within the FAR(p, m) framework, and ‘$\underline{\mathbf{g}}$’ represents a vector of optimal dimensions for a number of functional covariates. The theoretical contributions are twofold: first, deriving the distribution of the modified functional final prediction error, denoted as fFPE$\mathbf{X}(p,m,\underline{\mathbf{g}},\tau )$; second, using this derivation to establish formal criteria for optimal model selection. To empirically investigate the predictive performance of the proposed model, hourly temperature data from the NASA POWER project are considered, and day-ahead out-of-sample forecasts over a full annual cycle are computed. The forecasting performance of the proposed model is assessed against state-of-the-art models using different error summary metrics. The results show that functional models consistently outperform traditional time series and neural network-based approaches, with FAR$\mathbf{X}(p,m,\underline{\mathbf{g}},\tau )$ achieving superior predictive accuracy compared to FAR(p, m) and FARX(p, m, τ), thereby underscoring the efficacy of incorporating functional exogenous information in FTS modeling. Full article