Search Results (241)

Low-precision floating-point computation has become central to efficient artificial intelligence, yet its behavior for compact transformer-based hyperspectral imaging (HSI) models remains underexplored. In this work, we present a controlled comparative study of fully quantized training (FQT) and post-training quantization (PTQ) for pixel-wise foreign plastic object (FPO) classification in poultry hyperspectral data. Using a fixed state-of-the-art spatial–spectral transformer backbone, a common mixed-precision strategy, and identical training and inference protocols, we evaluate FP32, FP16, BF16, FP8, and NVFP4 across predictive performance, model compression, training efficiency, and inference efficiency. The results show that mixed-precision FQT remains highly robust across the tested precision spectrum, with all reduced-precision configurations staying within 0.63 percentage points of the FP32 baseline in overall accuracy while consistently outperforming PTQ at matched precisions. Across the evaluated formats, BF16 provides the closest accuracy to FP32, whereas FP8 offers a particularly favorable balance between accuracy preservation and reduced precision, while model compression increases progressively to $3.69\times$ under NVFP4. The computational benefits, however, are strongly workload dependent. Native FP8/FP4 hardware support does not automatically improve training throughput for this compact model at moderate workloads, and the larger training batches required to better utilize low-precision hardware can degrade predictive performance. In contrast, large-batch inference can effectively exploit FP8 and NVFP4 without affecting predictive accuracy. An ablation study further shows that selective retention of numerically sensitive modules in FP32 is essential for stable ultra-low-precision operation. Overall, the findings demonstrate that low precision is a viable but workload-dependent design choice for compact HSI transformers, with FQT providing greater accuracy robustness than PTQ and FP8, offering a favorable overall accuracy–efficiency trade-off. Full article

Polyethylene terephthalate (PET) is a common plastic widely used in food packaging, especially plastic bottles, and in fibers and textiles. The widespread use of PET and its intentional and unintentional release and disposal over the past few decades have placed significant pressure on the environment, necessitating the development of green, low-cost, and efficient recycling technologies to mitigate this impact. This article reviews recent advances in microwave-assisted catalytic depolymerization of PET. It begins by outlining the fundamental principles of PET materials science and depolymerization mechanisms. The article then reviews the historical evolution of this research and presents a benchmark comparison of different depolymerization methodologies. Finally, through a critical analysis and comparison of state-of-the-art approaches, the article identifies emerging trends and highlights promising directions for future research in the field. A comprehensive comparison of conventional and microwave heating methods is presented, indicating that catalyst-assisted microwave systems can shorten reaction times and achieve high product yields under optimized conditions. Key advantages and limitations are highlighted, and persistent challenges are discussed. Overall, this article surveys the latest progress in the chemical recycling of PET and provides critical insights for future research and further development of microwave-assisted catalytic PET depolymerization technology. Full article

Phthalate esters (PAEs), which are widely used as plasticizers in various products, are recognized as endocrine-disrupting chemicals (EDCs) that can affect female reproductive health. Numerous studies linked PAE exposure to several health hazards, including disruption of folliculogenesis, steroidogenesis, implantation, and pregnancy maintenance, contributing to complications such as miscarriage, preeclampsia, and preterm birth. In response to growing concerns regarding their toxicity, alternative plasticizers (APs) have been introduced, with particular attention given to the commonly used compounds, including di(isononyl) cyclohexane-1,2-dicarboxylate (DINCH), tris(2-ethylhexyl) trimellitate (TOTM), di(2-ethylhexyl) adipate (DEHA), and acetyl tributyl citrate (ATBC). While these alternatives are often presented as safer, recent studies have suggested that these compounds can also disrupt hormonal pathways and fertility regulation mechanisms. Both PAEs and APs have been shown to interact with key nuclear receptor systems, especially with peroxisome proliferator-activated receptors (PPARs) and steroid hormone receptors. These receptors are essential regulators of ovarian function, placental development, and pregnancy maintenance. Disruption of these signaling pathways, particularly in the placenta and fetal membranes, may contribute to altered inflammatory responses, impaired endocrine regulation, and increased risk of adverse pregnancy outcomes. This review provides a state-of-the-art synthesis of the current literature on PAEs and APs, focusing on exposure patterns, associations with female reproductive health and pregnancy outcomes, and the identification of dysregulated biological pathways. This review emphasizes the urgent need to limit plasticizer exposure, especially among women of reproductive age, in order to preserve reproductive health and prevent associated complications. Full article

Current mainstream deep learning techniques exhibit an over-reliance on extensive training data and a lack of adaptability to the dynamic world, marking a considerable disparity from human intelligence. To bridge this gap, Few-Shot Class-Incremental Learning(FSCIL) has emerged, focusing on continuous learning of new categories with limited samples without forgetting old knowledge. Existing FSCIL studies typically use a single model to learn knowledge across all sessions, inevitably leading to the stability–plasticity dilemma. Unlike machines that usually consolidate all categories into a single parameter space, cortical memory organization suggests that different types of knowledge can be distributed and organized across specialized cortical regions. Inspired by this organization principle, our paper aims to develop a method that learns independent models for each session. It can inherently prevent catastrophic forgetting. During the testing stage, our method integrates Uncertainty Quantification (UQ) for model deployment. Our method provides a fresh viewpoint for FSCIL and demonstrates the state-of-the-art performance on CIFAR-100 and mini-ImageNet datasets. Full article

The construction sector is under enormous pressure to reduce the overall environmental impact of its activities, which, for cement production alone, cause over 2 Gt/yr of CO₂ emissions (2022). Given the massive scale of concrete production, even minor components such as plasticizers, typically required in quantities of only a few percent of the dry cement mass, correspond to production volumes of many tens of Mt/yr, making their sustainability highly relevant. Lignin is the second most abundant renewable biopolymer on Earth and is currently used in concrete formulations mainly as a low-performance cement plasticizer. Despite this limited application, lignin represents an underexploited but highly attractive raw material for the development of environment-friendly concrete additives, owing to its abundance and availability from industrial biomass residues. The purpose of this review is to provide a specific and up-to-date overview of the various approaches for developing sustainable, high-performance concrete plasticizers based on lignins derived from industrial waste streams. The state of the art in the field is put into perspective by considering the various industrial sources of original technical lignins, the processes that led to their formation, and potential strategies for improving their characteristics. A quick overview of the patent history of the field complements the analysis of the scientific literature. An integral part of this work is a concise introduction to the fundamentals of cement chemistry and rheology, which are essential for understanding the technological requirements for effective concrete admixtures and for enabling cross-disciplinary research between the lignin and cement communities. Full article

Polymers play a central and multifaceted role in cultural heritage science, serving both as functional materials in conservation treatments, such as cleaning, consolidation, adhesion and protection, and as constituents of a wide range of historical artefacts, including paper and canvas, waterlogged wooden wrecks, musical instruments, and modern plastics used in art. Although numerous studies have examined the use of polymers in archaeology and cultural heritage conservation, most have focused on specific polymers, individual conservation treatments, or categories of artefacts. A comprehensive and integrated assessment of the multifunctional role of polymers, both as conservation materials and as constituents of heritage objects, remains lacking. The aim of this review is to provide a critical and comprehensive overview of natural and synthetic polymers in cultural heritage science, examining their applications in conservation treatments, their long-term stability and aging, and the challenges and opportunities associated with their preservation and sustainable use. This review examines the main classes of natural and synthetic polymers used in conservation, evaluating their mechanisms of action, performance, limitations, and long-term behavior across different applications. It also examines the chemical decomposition pathways and the resulting degradation phenomena occurring in polymeric materials, both as conservation products and as constituents of cultural artefacts, together with current stabilization strategies aimed at mitigating aging and deterioration. This review provides a critical appraisal of current challenges and future perspectives in cultural heritage conservation, highlighting emerging trends and research directions for the development of more effective, sustainable, and compatible polymer-based solutions for cultural heritage conservation. Full article

We present a unifying thesis which posits that the biological resolution of uncertainty is a fundamental adaptation to entropy’s negative impact on the highly ordered molecular structures required to maintain the living state. These molecular biological adaptations are highly conserved and play a critical role in the survival of the earliest multicellular organisms and the vertebrates thereafter. The imperative to reduce cognitive uncertainty is effected through the dopaminergic Medial Forebrain Bundle (MFB) Reward Prediction Error (RPE) mechanism, or its homologous equivalents, to compute a biological valuation of information. This hypothesis is supported by the central role of the MFB seeking system as the neural substrate of exploratory behavior that leads to the reduction of uncertainty when information is apprehended and cognitively assimilated. We define the human experience of intellectual beauty as the subjective emotional reward that is activated by the MFB seeking system. Accordingly, humans experience intellectual beauty when a high-entropy state of confusion is suddenly resolved into a low-entropy state of insight. In humans, the neuroanatomical substrates of inductive reasoning, inquiry, and the intellectual beauty to which they lead are present at birth. What develops postnatally is synaptic plasticity in the connections among these neurons that is activated in the loving didactic relationship that is established between mother and child during infancy. This dynamic is critically dependent on observational learning on the part of the child. It is supported by the joyful engagement and emotional support of the mother. This provides a paradigm of joy in learning that we propose is the developmental origin of intellectual beauty. This is the reinforcement that maintains inquiring behavior in the search for information that is needed to resist the adverse effects of entropy on life. This paper traces the continuous thread of uncertainty resolution from its phylogenetic origins in associative learning to the intuitive science of early childhood, and ultimately to the highest levels of human inquiry in science, as well as literary, musical and visual arts. The intuitive scientific method gives rise to the collective intelligence of groups, an evolved trait that likely contributed to the success of our hominin ancestors. At the societal level, this collective intelligence scales into the institutional working of markets, driving the macroeconomic price discovery of new information to counter entropy. Importantly, we compare the cost of information across the disparate domains of pharmaceutical drug discovery and the contemporary art market to demonstrate that the imperative to reduce uncertainty manifests as a universal, falsifiable mechanism for the “price discovery” of information. Full article

Die forging is a widely used production method for manufacturing high-strength components with high accuracy and good mechanical properties. Since the die forging process involves many complicated thermo-mechanical coupled field physical problems, such as large plastic deformation, high temperature, friction, and heat transfer, etc., experimental studies are difficult and expensive to perform. The numerical simulation method has become the main method of study and optimal design for the die forging process. This paper reviews the published papers on numerical simulation of the die forging process from 2016 to 2026, in a structured literature review of computational simulations dealing with die forging processes. The literature search was conducted using the Scopus and Web of Science databases. After screening and full-text assessment, 24 relevant journal articles were selected for this paper. The articles studied were analyzed in terms of finite element modelling strategies, constitutive and material models used, friction and thermal boundary conditions considered, and process parameters. Typical results obtained from the studies discussed in the paper include stress, strain, temperature, forging load, and metal flow. The current state-of-the-art research has evolved from simple metal-flow predictions to more complex thermo-mechanical models, and even optimization-based models. Most of the current studies are based on experimentally derived constitutive equations, as well as more complex friction and heat-transfer models. Furthermore, studies applying optimization methods (Taguchi methods, design of experiments, machine learning, artificial intelligence) are increasingly common. The growing interest in the digital twin concept and real-time process control is observed. However, experimental validation, thermal contact modelling, and simulation of stress, strain, temperature, and microstructure in one simulation remain key challenges. Full article

Background: Mitochondria are the primary organelles that regulate cellular bioenergetic metabolism and maintain homeostasis, providing essential structural support for optimal cell survival. Nonetheless, advancing age leads to cumulative damage to mitochondrial structure and functional integrity, which is a defining characteristic of biological aging and is closely linked to the emergence and progression of numerous age-related diseases, including neurodegenerative disorders, cardiovascular diseases, and metabolic disorders. Scope of review: This article offers a thorough summary and review of mitochondrial quality control (MQC), emphasizing numerous critical processes, including mitochondrial biosynthesis, dynamic remodeling (fusion and fission), and mitophagy. We thoroughly elucidate the molecular pathways that regulate MQC and demonstrate how age-related dysregulation precipitates cellular senescence, highlighting the transition from physiological maintenance to pathological malfunction, which ultimately culminates in cellular aging. Conclusions and implications: This study systematically elaborates the pathophysiological mechanisms in the field, comprehensively evaluates the clinical translational potential of targeting the MQC pathway, highlights the key objectives of “restoring mitochondrial plasticity and removing dysfunctional mitochondria”, and explores novel intervention strategies. The restoration of normal mitochondrial function in cells throughout aging is a very promising path for precision medicine therapeutics with great translational potential, according to recent state-of-the-art research. The development of novel therapeutic approaches to improve functional healthy mitochondria can effectively delay aging and reduce the rising global burden of age-related diseases. Full article

Traditional Android malware detection systems struggle to adapt to evolving threats without sacrificing performance on legacy families. To address this, we present ProtoMal, a dual-branch continual learning framework that achieves a fine-grained balance between stability and plasticity. The framework utilizes a frozen old branch for knowledge preservation and a trainable new branch for novel threat acquisition. A key contribution is our robust median-based prototype learning mechanism, which leverages centroids and outlier filtering to handle the high intra-class variability and label noise inherent in malware datasets. Experimental results across three large-scale benchmarks AMD, VirusShare, and VirusShareYears demonstrate that ProtoMal significantly curtails performance degradation and achieves highly competitive average accuracy. Most notably, the proposed framework demonstrates highly competitive model stability and yields robust anti-forgetting capabilities alongside current state-of-the-art incremental learning paradigms, maintaining particular resilience under severe concept drift. Full article

The deployment of lightweight object detectors on remote sensing edge platforms is severely constrained by the rigid trade-off between perception capacity and metabolic expenditure. To solve this fundamental challenge, we draw inspiration from the superior energy efficiency of the mammalian brain and the principles of connectomics to introduce CONERSLite. By emulating the dual mode synergy of biological neural systems, CONERSLite integrates a Compact Anatomical Backbone (CAB) representing the stable anatomical connectome and a Functional Connectome Router (FCR) that mimics the plasticity of the functional connectome. Our framework achieves a peak mAP of 82.35% on the DOTA-v1.0 dataset with only 28.3 M parameters and 195 G FLOPs, effectively establishing a new accuracy–efficiency Pareto frontier for remote sensing. On the HRSC2016 dataset, it reaches a state-of-the-art mAP of 98.62% while reducing the total parameter count by approximately 45% compared to high-precision optimized models like RTMDet. These results demonstrate that the application of connectomics principles provides a biologically grounded and highly efficient solution for resource-constrained remote sensing object detection. Full article

Plastic design enables efficient structural systems by exploiting controlled inelastic deformation and force redistribution. While mature in steel structures due to stable ductility and well-defined yielding, its extension to reinforced concrete (RC) remains challenging because cracking, stiffness degradation, confinement dependency, and progressive damage govern deformation capacity and collapse mechanisms. This paper presents a state-of-the-art review of optimal plastic design methodologies for RC structures by tracing the evolution from classical plasticity theory to modern damage-informed, reliability-oriented, and sustainability-driven formulations. A systematic and structured literature review of more than 90 peer-reviewed journal articles (1990–2025) was conducted using Scopus, Web of Science, and ScienceDirect. The selected studies are classified by structural system type, plastic analysis approach, constitutive modeling strategy, and strengthening technique, including CFRP and hybrid fiber systems, optimization framework, and uncertainty treatment. The review highlights how nonlinear elasto-plastic and damage–plasticity models improve the prediction of plastic hinge development, redistribution, and failure-mode transitions, and how metaheuristic optimization, topology optimization, surrogate modeling, and machine learning are increasingly used to manage discrete design variables and computational cost. Reliability-based methods (e.g., FORM/SORM and simulation) are shown to be essential for quantifying deformation-capacity uncertainty and ensuring consistent collapse-prevention performance. A comparative assessment of nine plastic design methodologies is also provided, identifying their core assumptions, limitations, and domains of applicability within a structured evaluative framework. Remaining challenges include robust deformation-capacity prediction, reproducible calibration of damage models, and integration of life-cycle sustainability criteria within reliability-constrained plastic optimization. Future research directions are proposed toward multi-objective reliability-based design, durability-informed plastic modeling, and hybrid physics-informed AI-assisted workflows. Full article

Amblyopia is increasingly conceptualized as a neurodevelopmental visual disorder that often arises from discordant binocular visual experience during early life and is associated with abnormal binocular interactions, interocular suppression, orientation-dependent developmental abnormalities in selected refractive phenotypes, and experience-dependent plasticity, consistent with a distributed-network perspective rather than a purely monocular acuity deficit. We performed a structured state-of-the-art narrative synthesis of peer-reviewed reviews, randomized controlled trials, and key mechanistic human studies indexed in PubMed/MEDLINE, Web of Science, and Scopus (1 January 2016–28 February 2026; last search 28 February 2026), prioritizing recent evidence from 2021–2026. Literature supports consideration of clinically trackable constructs beyond best-corrected visual acuity (BCVA), including quantified suppression/imbalance, binocular function, and functionally meaningful outcomes such as reading-related limitation and broader functional impact. Across established and emerging intervention classes, treatment effects are heterogeneous across ages and etiologies. Evidence is strongest for conventional penalization and selected active training-based approaches, whereas newer protocol-standardized approaches remain investigational and require prospective evaluation with transparent exposure/dose reporting. Based on these findings, we outline a clinically oriented, core outcome set for amblyopia and strabismus (COSAMS)-aligned framework that combines quantified binocular imbalance with multidimensional phenotyping and a hypothesis-driven, prospectively testable therapeutic model intended to structure (not replace) clinical decision-making. Priorities for precision-oriented amblyopia care include standardization of suppression metrics, adoption of core outcome sets, transparent reporting of ‘not measurable’ outcomes and missingness, and prospective validation of phenotype-driven, prediction-ready frameworks. Full article

Few-Shot Class-Incremental Learning (FSCIL) aims to sequentially learn new classes from very few labelled samples while preventing the forgetting of previously acquired knowledge, which has important practical value for remote sensing scene classification (RSSC). Recent studies have shown that applying a Vision Transformer (ViT) pre-trained on natural image datasets to FSCIL tasks can achieve significantly superior performance. Nevertheless, a substantial domain distribution gap exists between natural images and remote sensing images, which leads to severe performance degradation when such models are directly transferred to RSSC. To address the domain gap alongside FSCIL’s inherent stability–plasticity dilemma and overfitting under data scarcity, we propose a Dynamic Expansion Mixture-of-Experts with Pre-trained Vision Transformer (DEM-ViT) framework. Specifically, to alleviate the domain discrepancy, DEM-ViT incorporates an Adapter-Based Mixture-of-Experts (ABMoE) module, which captures the diverse visual patterns of remote sensing scenes through feature reconstruction in the representation space and collaborative learning among multiple experts. Furthermore, to address the stability–plasticity dilemma in FSCIL, we propose a Dynamic Expert Expansion (DEE) strategy, which progressively expands the model capacity along the incremental sessions. DEE provides sufficient space for learning new knowledge while mitigating the forgetting of previous knowledge. In addition, we propose a Semantic-Guided Feature Alignment (SGFA) method to reduce the risk of overfitting under data-scarce conditions. SGFA leverages textual information to construct robust text prototypes and uses them to calibrate the visual feature space. Extensive experiments across three benchmarks indicate that our framework exhibits highly competitive performance compared with state-of-the-art methods. Full article

Automated pavement distress classification using high-resolution Unmanned Aerial Vehicle (UAV) imagery is pivotal for intelligent transportation systems. However, long-term UAV monitoring faces a continuous stream of evolving distress types and changing remote sensing background textures, necessitating Class-Incremental Learning (CIL) capabilities. Existing methods struggle to balance stability and plasticity, especially under the severe storage limitations typical of local edge stations in air–ground collaborative systems. This data scarcity leads to catastrophic forgetting and confusion among fine-grained distress categories. To address these challenges, we propose a data-efficient approach named Analytic Geometric Alignment (AGA). Our framework mainly consists of three key components. First, to overcome the optimization gap between the feature extractor and the fixed geometric target, we introduce a Subspace-Aware Analytic Initialization (SAI) that computes a closed-form projection to instantly align the feature subspace with the ETF manifold before each task training. Second, on this aligned basis, a Decoupled Geometric Adapter (DGA) is incorporated to facilitate continuous non-linear adaptation to complex aerial textures. Finally, for stable incremental training, we design a Memory-Prioritized Regression (MPR) loss to enforce tighter geometric constraints on replay samples, significantly enhancing model stability. Extensive experiments on the UAV-PDD2023 dataset demonstrate that AGA significantly outperforms state-of-the-art methods, showcasing excellent robustness and data efficiency. Full article