Search Results (651)

Deep learning systems deployed in regulated settings require explanations that are accurate and stable under nuisance transformations, yet classical post hoc transition matrices rely on fidelity-only fitting that fails to guarantee consistent explanations under spatial rotations or other group actions. In this work, we propose Equivariant Transition Matrices, a post hoc approach that augments transition matrices with Lie-group-aware structural constraints to bridge this research gap. Our method estimates infinitesimal generators in the formal and mental feature spaces, enforces an approximate intertwining relation at the Lie algebra level, and solves the resulting convex Least-Squares problem via singular value decomposition for small networks or implicit operators for large systems. We introduce diagnostics for symmetry validation and an unsupervised strategy for regularization weight selection. On a controlled synthetic benchmark, our approach reduces the symmetry defect from 13,100 to 0.0425 while increasing the mean squared error marginally from 0.00367 to 0.00524. On the MNIST dataset, the symmetry defect decreases by 72.6 percent (141.19 to 38.65) with changes in structural similarity and peak signal-to-noise ratio below 0.03 percent and 0.06 percent, respectively. These results demonstrate that explanation-level equivariance can be reliably imposed post-training, providing geometrically consistent interpretations for fixed deep models. Full article

General Matrix Multiplication (GEMM) is a fundamental computational kernel in scientific computing, serving as the foundation for numerous complex tasks. However, in practical applications, the performance of GEMM is often constrained by irregular matrix dimensions and the diversity of hardware architectures. In particular, when processing small and irregular matrices, GEMM typically exhibits reduced computational efficiency. To address these challenges, this paper proposes a GEMM acceleration method based on an adaptive core grouping strategy. The method consists of two key components: a core grouping mechanism that alleviates workload imbalance among multi-core CPUs, and an adaptive block partitioning algorithm that dynamically selects optimal tiling schemes according to the matrix dimensions, achieving both load balance and cache-friendly data access. Experimental results on the Kunpeng CPU platform demonstrate that the proposed method achieves significant performance improvements compared to the Kunpeng KML math library, reaching a peak acceleration of up to 2.1× and an average speedup of 1.64×. These results validate the effectiveness and efficiency of the proposed approach in handling small and irregular matrix computation scenarios. Full article

Accurate biomass quantification is important for evaluating growth kinetics and performance of microalgal and microalgal–bacterial wastewater treatment systems. However, small-scale studies frequently encounter methodological limitations due to low biomass concentrations, limited sampling volumes, and/or interference from non-biotic solids in complex wastewaters. This work adopts a two-fold approach: (i) a concise review of current biomass quantification methods for bench-scale systems, and (ii) an experimental evaluation of a gravimetric protocol for complex wastewaters. The review discusses commonly applied techniques, highlights their strengths and weaknesses, and identifies research gaps in data comparability and reproducibility. The laboratory investigations evaluated the effects of key factors, namely culture volume (250 mL to 1 L), test aliquots (2.5 mL to 10 mL), and the absolute weight of total suspended solids (3.43 g to 14.5 g) on total suspended solids measurements. Aliquots containing <5 mg total suspended solids produced statistically significant variability, whereas reliable and reproducible results were obtained when >8–10 mg absolute total suspended solids per aliquot was present. In complex wastewater matrices, approximately 18% of total suspended solids consisted of non-volatile solids, demonstrating that the method can systematically over-estimate true dry cell weight in microalgal–bacterial systems. The findings emphasized the need for procedural standardization. Finally, a practical gravimetric protocol is proposed for both axenic and consortium-based small-scale studies dealing with complex wastewater, providing an evidence-based roadmap for obtaining more reliable biomass estimations. Full article

Urbanization is a major and rapidly expanding form of land-use change worldwide and is one of the main drivers of the decline in arthropod diversity. Within urban matrices, remnants of natural or semi-natural habitats serve as important refuges for native organisms. However, these urban fragments are typically small, isolated, and strongly affected by various forms of disturbance. Therefore, connectivity among urban remnant patches may enhance population persistence and resilience. Increased tendencies to explore novel environments, tolerate human disturbance, and exploit unpredictable resources can be advantageous in urban environments. Accordingly, in this study of a flightless ground beetle species, we hypothesized that individuals from urban habitats—especially from isolated ones—would be bolder and more exploratory than their rural conspecifics, that sexes would differ in behavior, and that these behaviors would be temporally consistent, indicating animal personality. Activity-, exploration-, and boldness-related behavioral traits were significantly repeatable, providing evidence for animal personality, particularly in females and rural beetles. Contrary to our hypothesis, no behavioral differences were detected between rural and urban individuals. Furthermore, no significant sex-dependent differences in behavior were observed. Overall, our findings highlight the importance of species-specific traits and ecological context in shaping behavioral variation. Full article

Despite offering rich complementary information, multimodal remote sensing images collected by diverse sensors increase the computational burden in clustering. To alleviate this issue, we devise an efficient multimodal clustering approach (MCSM) on superpixel manifolds formed by superpixel segmentation. The MCSM jointly learns cluster representation of all modalities and a consensus cluster membership graph that fuses the multimodal representation to yield clusters. To capture the local geometric structure of the superpixel manifolds, the optimization is constrained by manifold regularization of the consensus graph. In contrast to vanilla multiview subspace clustering techniques, the proposed approach does not rely on spectral clustering, and only involves element-wise product and multiplication on small-scale matrices. In addition, we prove that the MSCM is a special case of classic low-rank subspace clustering models, providing a perspective for understanding the learned cluster graphs. Extensive experiments are conducted on three popular multimodal remote sensing datasets, showing that the proposed method achieves competitive clustering performance compared to state-of-the-art methods, and significantly outperforms the latter in computational efficiency. Full article

As small modular reactors (SMRs) evolve towards longer lifespans, autonomous operation, and high reliability, the accuracy and reliability of sensor data are crucial for ensuring the safe operation of nuclear power systems. To improve the accuracy of multi-source sensor fault detection and localization in small reactors, this paper proposes a multi-scale cross-correlation-based convolutional autoencoder (MSCC-CAE) framework. First, multiple sensor cross-correlation matrices are constructed across multiple time scales to explicitly characterize the dynamic coupling relationships between heterogeneous sensors. These multi-scale correlation features can effectively capture both short- and long-term dependencies among sensors. Then, a convolutional autoencoder is used to compress and reconstruct the correlation matrix, thereby learning low-dimensional discriminative representations for fault detection. To enhance the stability and generalization of the proposed framework, a multi-strategy improved quantum particle swarm optimization (MSQPSO) algorithm is proposed to adaptively optimize key network hyperparameters. Finally, the proposed method was validated using data from an SMR simulation model. Experimental results demonstrate that the proposed MSCC-CAE achieves a fault detection accuracy of 98.21%, outperforming CNN and conventional CAE models by 15.17 and 12.04 percentage points, respectively. The localization accuracy reaches 97.12%. These results verify the effectiveness and superiority of the proposed framework for intelligent sensor fault detection in the SMR system. Full article

Low-rank sparse decomposition models have become the mainstream optimization framework for multiframe infrared small target detection. Existing low-rank matrix decomposition approximations typically pre-decompose infrared videos into the product of two low-rank matrices to capture the background’s low-rank characteristics. However, such approximations are not optimal and often result in suboptimal background recovery. To achieve more accurate low-rank recovery, we exploit the intrinsic relationship between low-rank matrices and their generalized inverse matrices, thereby improving conventional decomposition approximations. Moreover, to address the high computational cost of applying low-rank and sparse decomposition models to multi-frame infrared videos, we introduce a robust column-row (CUR) decomposition to accelerate the iterative process, thereby significantly improving computational efficiency. The experimental results show that the proposed method achieves fast detection of small targets in infrared videos while maintaining competitive detection performance. Full article

pH-responsive indicator films for intelligent food packaging applications are based on the embedding of a natural or synthetic dye in a polymeric substrate, preferably biobased and biodegradable. Although natural colorants like anthocyanins were extensively investigated in this respect, nature-inspired synthetic flavylium compounds could represent an alternative based on their higher stability. In this work, five novel synthetic 4′-aminoflavylium derivatives with different substitution patterns in the benzopyrylium core (compounds 1–5) were synthesized and characterized. Polyvinyl alcohol (PVA), as well as chitosan–PVA and chitosan–starch blends, were used to prepare pH-responsive indicator films having inserted each of the synthesized flavylium dyes or a natural onion peel extract. The PVA films with compounds 1 and 3, and the PVA–chitosan film with compound 1, exhibited antioxidant activity, highlighting their potential for active packaging applications. All indicator films showed pH responsiveness in the range of 2 to 12 and were subsequently tested in contact with the packaging atmosphere or in direct contact with pork and fish meat, at different temperatures (4 °C, 20 °C, and 40 °C) for 24 h to assess their colorimetric response to progressive spoilage. Although the differences were small, the films with the 7-hydroxy-4′-aminoflavylium derivative exhibited the earliest and most intense color change during storage of meat, starting from direct contact at 4 °C for 24 h, being able to identify the initial stages of meat spoilage, while the performance of the dihydroxy-substituted derivative was attenuated by incorporation in polymer matrices. This behavior was comparable to that of onion peel extract, but the synthetic flavylium derivative was more stable. The results can provide new opportunities for intelligent food packaging applications using biopolymer indicator films with 4′-aminoflavylium derivatives. Full article

Background: Disability progression in multiple sclerosis (MS) is increasingly recognized as a consequence of large-scale brain network disruption rather than isolated regional damage. Although diffusion tensor imaging (DTI) is the reference method for assessing structural connectivity, its limited availability restricts widespread clinical application. There is therefore a critical need for alternative approaches capable of capturing network-level alterations using routinely acquired MRI data. Objective: This study aimed to determine whether synthetic structural connectivity matrices derived from standard regional volumetric MRI can capture clinically meaningful network alterations in MS and predict subsequent functional progression, particularly upper limb decline. Methods: Regional brain volumetry was obtained from routine T1-weighted MRI using an automated, clinically approved volumetric pipeline. Synthetic structural connectivity matrices were generated by integrating principles of structural covariance, distance-dependent connectivity, and disease-specific vulnerability patterns. Graph-theoretical network metrics were extracted to characterize global and regional topology. Machine learning models including logistic regression, support vector machines, random forests, and gradient boosting were trained to predict clinical progression defined by worsening on the 9-Hole Peg Test. Dimensionality reduction was performed using principal component analysis, and model performance was evaluated using balanced accuracy, AUC-ROC, and resampling-based validation. Feature importance analyses were conducted to identify network vulnerability patterns. Results: Synthetic connectivity networks exhibited biologically plausible properties, including preserved but attenuated small-world organization. Global efficiency showed a strong inverse correlation with disability severity (EDSS). Patients with clinical progression demonstrated marked reductions in network integration and segregation, alongside increased characteristic path length. Machine learning models achieved robust prediction of upper limb functional decline, with ensemble-based methods performing best (balanced accuracy > 80%, AUC-ROC up to 0.85). A limited subset of connections accounted for a disproportionate share of predictive power, predominantly involving frontoparietal associative networks, thalamocortical pathways, and inter-hemispheric connections. In a longitudinal subset, network-level alterations preceded measurable clinical deterioration by several months. Conclusions: Synthetic structural connectivity derived from routine volumetric MRI captures clinically relevant network-level disruption in multiple sclerosis and enables accurate prediction of functional progression. By bridging network neuroscience with widely accessible imaging data, this framework provides a pragmatic alternative for connectomic analysis when diffusion imaging is unavailable and supports a network-based understanding of disease evolution in MS. Full article

Matrix effects (ME) during LC-ESI-MS analysis are a commonly acknowledged issue for a variety of matrices and analytes. Although sample preparation techniques are steadily evolving to reduce ME, the complexity and variability of the urine matrix remain a challenge, especially for multi-analyte methods. To investigate the extent of ME implications on method performance and quantification, we used stable isotope-labelled standards (SIL-IS) of 11 mycotoxins to evaluate the magnitude and variability of ME in urine samples from two cohorts: Bangladeshi adult women (n = 50) and Swedish children of both sexes (n = 340). Significant ME differences were observed between the two cohorts for eight of the 11 mycotoxins. Additionally, intra-cohort ME variability turned out to be very high with interquartile ranges (IQR) above 15% for 14 out of 22 analyte-cohort combinations. Maximum IQR values were observed for sterigmatocystin in the Bangladeshi cohort (318%), strongly impacting quantitative results obtained with matrix(-matched) calibration. Further experiments on a small German cohort of four subjects, each providing four to five urine samples, revealed high variability of ME within each individual. Factors influencing ME were investigated, showing little to no impact of sex and a moderate impact of age for some analytes in the Swedish cohort. Nonetheless, especially the more polar analytes, showing stronger signal suppression, demonstrated clear correlation of ME with density and creatinine concentration of the urine samples. As a result, urine samples with very high or low density or creatinine values require careful handling in regard to sensitivity or quantification errors when matrix(-matched) calibration without SIL-IS is applied. Full article

We revisit the black-hole information problem from the viewpoint of a population–coherence decomposition of density-matrix purity. Building on a previously developed formalism for n-dimensional density matrices, we characterize each state by a normalized global purity index and two complementary indices, which quantify the contributions of level populations and coherences. This yields a simple quadratic relation and a geometric representation in a “population–coherence plane”, where different routes to purity can be distinguished. In the two-level case, we construct explicit families of states with identical spectra and global purity but opposite internal structure, realizing population-dominated and coherence-dominated routes. We then apply this framework to a standard Page-type evaporation model without an explicit Hamiltonian, in which a black hole and its Hawking radiation form a bipartite pure state with varying Hilbert-space dimensions. Using known results for typical reduced states in large dimensions, we analyze the behavior of population and coherence components of purity along the evaporation process. Under the physically motivated requirement that, in this energy-free setting, the radiation populations remain nearly uniform in the chosen basis, we show that the late-time recovery of purity must be coherence-dominated: the global purity of the radiation approaches unity while the population index stays small and the coherence index carries essentially all the purity. Full article

Up to now, most hydrogel-related studies have been devoted to the preparation of drug-containing macromolecular gels via the introduction of polymer matrices, together with the clarification of their assembly mechanisms and biomedical applications. In contrast, studies concerning the design of small-molecule gel systems remain relatively limited. As gel research progresses, drug small-molecule hydrogels have attracted growing interest for formulation development. This study investigated whether designing a small-molecule hydrogel could serve as an effective solubilization approach for sulindac (SUL)—a nonsteroidal anti-inflammatory drug clinically restricted by its poor aqueous solubility. Then, a SUL small-molecule hydrogel was prepared by straightforward mixing of SUL with biologically safe meglumine (MEG) in a minimal volume of deionized water, which exhibited a characteristic three-dimensional network structure and favorable viscoelastic properties. The characterization and simulation results indicated that the hydrogel formation was contingent upon the SUL-MEG miscibility, dissolution-aggregation equilibrium and intermolecular self-assembly. Consequently, the resulting SUL-MEG hydrogel exhibited 546 times higher solubility compared to the pure SUL. Meanwhile, the SUL-MEG hydrogel demonstrated superior release kinetics and supersaturation capacity, characterized by rapid attainment of peak concentrations and sustained supersaturated release. These enhanced performances were attributed to the high-energy state of the hydrogel itself and the molecular complexation between SUL and MEG. In conclusion, this study presents a feasible formulation strategy for overcoming the poor water solubility of insoluble drugs through the development of small-molecule hydrogel formulations. Full article

The selective detection of microplastics (MPs) in aquatic environments is hindered by particle size diversity and matrix-induced interferences. This study reports an excitation–emission matrix (EEM) fluorescence sensing platform using polyamide-derived carbon quantum dots (PACQDs; 0.5–2.6 nm) for the size- and concentration-resolved detection of polyethylene terephthalate MPs (PETMPs). PACQDs exhibited a pronounced fluorescence “turn-off” response upon PETMP interaction, governed by particle size (10–149 μm) and loading (4–8 g L⁻¹). Small PETMPs (10 μm) followed linear Stern–Volmer behavior, achieving a detection limit of 1.67 mg L⁻¹ in deionized water. Conversely, larger particles induced non-linear optical effects, including scattering-driven enhancement and inner-filter effects. Multivariate analysis using PCA and PARAFAC resolved three distinct components associated with surface-state quenching, scattering-mediated redistribution, and surface area-driven binding. Component-specific scores confirmed that PACQDs are most sensitive to small PETMPs, while larger particles primarily introduce optical interference. Selectivity tests showed distinct discrimination of PETMPs over polyamide and polypropylene. In tap water, significant matrix effects were corrected via matrix-matched calibration, achieving recoveries within 80–120%. This study establishes EEM-based multivariate fluorescence as a mechanism-informed strategy for PETMP sensing, highlighting the robust applicability of PACQDs for monitoring small PETMPs in real-world water matrices. Full article

Aflatoxin B1 (AFB₁) is one of the most toxic fungal secondary metabolites. High-sensitivity and rapid detection of AFB₁ is crucial for safeguarding consumer health, reducing post-harvest food losses, and promoting agricultural trade. Here, we developed a magnetic enrichment–catalytic lateral flow immunochromatographic assay (E-C-LFIA) for quantitative AFB₁ detection. The approach couples immunomagnetic capture and enrichment with carboxylated magnetite (Fe₃O₄) nanozyme probes and post-assay peroxidase-like catalysis of the H₂O₂–TMB system to enhance colorimetric readout. Compared with conventional LFIA performed without magnetic enrichment or catalytic amplification, E-C-LFIA achieved a visual detection limit of 0.05 μg/L for AFB₁, corresponding to a 20-fold improvement in sensitivity. The quantitative limit of detection (LOD, 3σ) was 0.023 μg/L, representing a 14.8-fold improvement in sensitivity. The method was demonstrated for AFB₁ screening in representative cereal- and nut-based matrices (rice, corn and peanut). Overall, E-C-LFIA provides a sensitive, rapid, and equipment-light option for on-site AFB1 screening and offers a transferrable strategy for other small-molecule contaminants. Full article

Unmanned Aerial Vehicle (UAV) inspection of transmission lines faces two primary challenges when detecting and analyzing components or defects in videos or images: poor performance in detecting small objects, and interference from complex backgrounds. To enhance defect detection under such cluttered conditions, this paper introduces an improved YOLO-based model, termed Transmission Line Defect Detection–YOLO (TLDD-YOLO), which jointly optimizes feature representation via a Dual-Branch Guided Attention (DBGA) mechanism and a Spatial Offset Attention Module (SOAM). DBGA employs a dual-branch structure to extract high-frequency spatial details and channel-wise semantic information, thereby guiding the backbone network to preserve the critical edge and texture features of small objects, mitigating detail loss during downsampling. SOAM utilizes a lightweight offset generation network to produce spatial offset matrices, and dynamically adjusts feature distributions through offset-guided spatial alignment, enabling feature contours to better conform to object shapes while reducing interference from complex backgrounds. The experimental results on a self-constructed transmission line inspection dataset demonstrate that TLDD-YOLO achieves 57.1% mAP, 83.8% mAP₅₀, and 36.1% mAP_s. Compared with the baseline model, the proposed method improves mAP, mAP₅₀, and mAP_s by 1.8%, 1.8%, and 7.7%, respectively, confirming its effectiveness for small object detection in UAV-based transmission line inspection. Full article