Evaluating the adaptation rates for esophageal cancer: Impact of the setup error contribution and of the dosimetric threshold.

Exploiting deep network with adaptive learning rate framework to recognize scattering vortex beam OAM mode

Transfer learning and explainable AI for heating load forecasting: A large-scale benchmark with SHAP-based static features

Conventional methods for treating potato starch processing wastewater (PSPW) are challenged by issues such as poor adaptability and high idle rates. Returning wastewater to the field, as a resource utilization strategy aimed at reducing pollutants and recovering nutrients, has emerged as a preferred approach for PSPW treatment. However, the unclear safe carrying capacity of soil in this process poses potential ecological risks. To scientifically evaluate the impact of repeated PSPW irrigation on soil safety performance, this study employed a soil infiltration system (SIS) to simulate the wastewater infiltration process. The results demonstrated that the SIS achieved average removal rates of 74.38%, 59.68%, 45.10%, and 77.94% for chemical oxygen demand (CODCr), total nitrogen (TN), ammonia nitrogen (NH4+-N), and total phosphorus (TP), respectively. Furthermore, multiple PSPW infiltrations led to weakly alkaline soil conditions and significant accumulation of organic matter, nitrogen, and phosphorus in the topsoil. Concurrently, the infiltration process altered the microbial community structure and reduced microbial diversity in the surface soil. The composition and function of the microbial community varied significantly with soil depth, with the community in the subsoil resembling that in the original, untreated soil. This study elucidates the patterns of pollutant removal during PSPW infiltration, the response of soil physicochemical properties, and the changes in microbial community structure at different soil depths. It provides a scientific basis for assessing the ecological impact of PSPW re-irrigation on soil and supports the potato industry in Northwest China in achieving green development goals.

The SystemC language, with its higher level of abstraction, plays a critical role in facilitating hardware/software co-design and architecture exploration. However, as most hardware models are predominantly written in Verilog and translating between SystemC and Verilog remains a challenge, an efficient and reliable tool for translating between these two languages is essential to streamline system development. This article proposes SCAV, a bidirectional translator between SystemC and Verilog, which breaks these limitations. SCAV provides a fully automated solution for translating both SystemC to Verilog and Verilog to SystemC, leveraging a translation framework with front-end/back-end separation. Additionally, SCAV incorporates an Abstract Syntax Tree (AST) filter, optimizing the translation process by filtering out invalid content. The experimental results demonstrate that SCAV achieves a 100% adaptation rate for Verilog and a 98% adaptation rate for SystemC, with 100% accuracy in both directions. Furthermore, SCAV outperforms existing tools, delivering a minimum speedup of 18% across various test cases.

Kiwifruit gray mold (caused by Botrytis cinerea) causes severe post-harvest losses. Traditional manual disease identification is inefficient, and post-harvest control measures lack systematic comparison. The aim of this study is to combine the residual network ResNet 18 model (ResNet 18-A) using Adam optimizer (which can maintain the adaptive learning rate of each parameter separately) with meta-analysis methods for preharvest diagnosis of kiwifruit gray mold and screening of effective post-harvest control measures, providing reference for scientific management of kiwifruit orchards. In terms of disease diagnosis, the disease recognition model based on ResNet 18 achieved an accuracy of 98.70% on a dataset of 9702 live images, meets the routine diagnostic needs. In terms of screening for efficient post-harvest prevention and control measures, the meta-analysis results showed that the treatment of soaked in MT solution for 3 min and stored at 22-24 °C with a relative humidity of 85%-95% (E1/E2), and the treatment of sprayed with Bacillus tequilensis KXF 6501 fertilization broth (1 × 107 CFU mL-1) has a good comprehensive effect on various kiwifruit quality indicators. The combination of ResNet 18 and meta-analysis can effectively improve the management efficiency of kiwifruit gray mold. Both models have great potential in achieving efficient disease diagnosis and targeted post-harvest prevention and control measures selection, which can help optimize disease management strategies for horticultural crops. © 2026 Society of Chemical Industry.

Malaria is one of the major global public health issues. An estimated 282 million malaria cases occurred worldwide in 2024, and the overall prevention and control progress has stagnated or even reversed in some regions. Mass drug administration (MDA), as a potential strategy to accelerate malaria elimination, has regained attention. This paper reviews the evidence base, controversial focuses, and application strategies of MDA in malaria prevention and control. It aims to promote its scientific application in the elimination phase. MDA plays an important role in malaria prevention and control. However, this strategy is accompanied by core limitations such as long-term drug resistance risks, insufficient implementation sustainability, and a high failure rate of regional adaptation. It also faces challenges from multiple common malaria species, as well as the newly discovered Plasmodium knowlesi. We therefore propose an “MDA+” collaborative strategy integrating vaccines, digital monitoring, and cross-border cooperation, so as to optimize resource allocation, achieve full coverage control over various malaria parasites, and advance the global malaria elimination process.

This study addresses a critical methodological gap in evaluating building envelope performance in hot, arid climates, the overreliance on annual energy indicators, which fail to capture transient thermal behavior during peak-load periods. In such environments, instantaneous heat gains, their intensity, and temporal distribution are decisive factors for cooling demand, occupant comfort, and grid stability. To overcome this limitation, a dynamic evaluation framework—the Thermal Adaptation Rating (TAC) system—is proposed. TAC integrates three interrelated indices—peak temperature reduction (ΔT_peak), relative peak cooling load reduction (ΔP_peak, %), and peak thermal delay (Δt_delay), representing thermal damping, load intensity mitigation, and temporal redistribution, respectively. A typical residential building in Karbala was modeled in DesignBuilder using the EnergyPlus engine, with inputs documented and calibration performed against real consumption data following ASHRAE standards (MBE and CV(RMSE)) to ensure reliability. The study examined advanced envelope systems, including thermochromic glass (TG), phase-change materials (PCMs), aerogel materials (AMs), and hybrid combinations. Results revealed that while AM achieved the greatest annual energy savings, its impact on instantaneous cooling load was limited. PCM, by contrast, effectively mitigated and delayed peak loads, enhancing thermal comfort (PMV/PPD). Hybrid systems, particularly TG-PCM, delivered the most balanced performance, simultaneously reducing peak cooling load and shifting its occurrence to reshape the cooling demand curve during critical periods. These findings demonstrate that annual indices alone are insufficient for evaluating envelope performance in extreme climates. Peak-condition analysis, expressed in terms of instantaneous cooling load, as operationalized through TAC, provides a more accurate representation of thermal behavior and offers a practical tool to guide envelope design decisions in hot, dry regions.

We compare the Heterogeneous Autoregressive (HAR) model with a novel Continuous Memory System (CMS) for forecasting realized volatility. CMS employs 12 exponential moving averages with adaptive decay rates modulated by learned, level-specific shock sensitivities through a rank-1 gating mechanism. The response of each memory level to volatility shocks is governed by an optimized sensitivity parameter that determines whether the level accelerates or decelerates during turbulent periods. Using 1,234 daily observations from February 2021 to January 2026, we estimate the model through bounded constrained optimization and compare its performance with that of the parsimonious HAR benchmark. CMS learns a surprisingly intuitive pattern in how different time horizons respond when markets become turbulent. Short-term memory reacts aggressively to volatility spikes, updating rapidly to capture sudden regime shifts. Medium- to long-term memory behaves in the opposite way, slowing down sharply during stress to preserve a stable baseline, with the strongest dampening occurring at horizons of roughly 4 to 6 weeks (levels 10 and 11). This creates an asymmetric response pattern: high reactivity at short horizons and strong stabilization at medium to long horizons. Notably, the model discovers this structure automatically from the data, without being explicitly designed to behave in this way, and the resulting pattern aligns closely with financial intuition about how different forecast horizons should weight past information during volatile periods. The sole exception is the 60-day horizon (level 12), which exhibits a large positive sensitivity. This may reflect distinct very long-term dynamics, or it may be an overfitting artifact, so it should be interpreted with caution. Despite its theoretical appeal, CMS underperforms in practice. Its out-of-sample forecasting error is 30% higher than that of HAR, even though it fits the training data extremely well. This is consistent with the classic problem of overfitting, in which a model captures historical patterns too closely and then fails to generalize well to new observations. The added complexity of CMS, with 25 tunable parameters versus HAR’s 4, appears to be a liability rather than an asset in a limited-sample setting. The sophisticated level-specific shock responses also provide almost no forecasting improvement, only 0.13%, over a simpler uniform-gating specification, while also creating numerical instability at very short horizons. Ultimately, HAR’s simplicity is a strength: fewer parameters leave less room for overfitting, making the model more reliable out of sample. At the same time, the response patterns learned by CMS, particularly how different forecast horizons adjust to volatility shocks, provide useful economic intuition that may help guide the design of better hybrid models in future work.

Dynamic task offloading strategy in mobile edge computing using meta-reinforcement learning with adaptive learning rate adjustment

Studies over the past 3 decades have shown decreased motor adaptation with age. The most widely supported theory for this change proposes that older adults fail to successfully implement cognitive strategies to facilitate performance. However, increased movement variability may also affect adaptation with age, but this has to date remained unstudied. Here, we examine whether age-related increases in movement variability influence adaptation. Healthy older adults (N = 48) exhibited significantly higher levels of baseline movement variability than younger adults (N = 60) in a visuomotor adaptation task, with variability increasing modestly with advancing age. Across all participants, greater movement variability correlated with slower rates of adaptation and smaller aftereffects. In our study, we also manipulated perturbation magnitude and sensory feedback in the task: such manipulations should facilitate participants applying cognitive strategies to the larger, more identifiable perturbation. Altering perturbation magnitude and sensory feedback did not affect the relationship between age and measures of adaptation performance, suggesting that adaptive changes with age cannot be fully accounted for by changes in the generation or the application of cognitive strategies in this task. While our results are correlational, they suggest instead that increased movement variability may contribute to the age-related decline in visuomotor adaptation. (PsycInfo Database Record (c) 2026 APA, all rights reserved).

Problem: Controlling nonlinearly parameterized systems with unknown disturbances remains challenging because classical adaptive approaches rely on separation-of-variables and reparameterization techniques, leading to increased parameter dimensions, conservative stability bounds, and implementation complexity. Objective: This paper develops a data-driven predictive fuzzy adaptive control (DD-PFAC) framework that eliminates the need for separation techniques while achieving superior tracking performance and formally certified stability. Novelty: The key innovation is a two-layer architecture. Layer 1 provides direct fuzzy approximation of composite nonlinear functions (system dynamics plus disturbance bound) without parameter reparameterization, reducing parameter complexity from O(qn) to O(nN). Layer 2 employs Hankel matrix-based predictive optimization to adaptively tune both control gains ci(k) and adaptation rates γi(k) online using 80–150 recent input–output samples. Methodology: A Lyapunov function augmented with a prediction-error term is used to prove uniform ultimate boundedness of all closed-loop signals. A projection-based recursive least-squares algorithm updates the gain parameters online while guaranteeing ci(k)≥cmin>0 at all times. Results: Comparative simulations demonstrate 31.4% reduction in integral square error, 27.8% reduction in mean absolute error, and 37.4% reduction in steady-state error versus traditional adaptive fuzzy control. A four-group ablation study confirms that adaptive gain scheduling contributes 27.7% and predictive compensation contributes 6.5% to the total MAE improvement. Robustness tests validate consistent 28–32% performance advantage across sinusoidal, pulse, step, and large-disturbance scenarios.

Heat shock factor-1 (HSF-1) plays a crucial role in orchestrating stress responses across diverse organisms and disease conditions. Here, we investigate how the HSF-1 signaling pathway influences the degradation of toxic proteins and neuropathological changes in the Caenorhabditis elegans model of amyotrophic lateral sclerosis (ALS). We found that overexpressing HSF-1 improves locomotor ability and increases the survival rate of ALS C. elegans. Moreover, we observed a deceleration of motor neuron degeneration, demonstrating the protective effect of HSF-1 on neurodegenerative processes. Transcriptomic analysis revealed notable changes in genes associated with autophagy and neurodegeneration, underscoring HSF-1's critical involvement in ALS pathology. In addition, metabolomic profiling further highlighted the involvement of this pathway in metabolic reprogramming. Overall, our study underscores the critical role of the HSF-1 signaling pathway in improving survival rate, movement velocity, cellular integrity, and metabolic adaptation, providing new insights into the mechanisms underlying ALS and potential targets for therapeutic intervention.

Cyber attacks and limited communication resources are two important problems in cyber–physical dc microgrid (MG). This article proposes an edge-event-triggered-based distributed resilient model-free adaptive control (MFAC) method for the dc MG suffering multipattern deception attacks to achieve secure voltage restoration and proportional current sharing. It is the first time to construct the distributed MFAC framework that can ensure the consensus of input signals, rather than the output bus voltage, to meet the fundamental requirement for distributed secondary controller design in the dc MG. Second, to deal with the multipattern deception attacks characterized by stochastic and successive variables, a novel data-driven predictive algorithm with an adaptive decay rate is developed to compensate for the polluted input signals, thus mitigating the negative impact of the attacks on the system. Moreover, an edge-event-triggered mechanism (EETM) with a switch-adjustable threshold parameter is designed to reduce communication frequency between any pair of adjacent DGs. Through rigorous mathematical analysis, the convergence of the bus voltage tracking errors is proved, and some comparison experiments are offered to verify the theoretical results.

Federated learning (FL) enables collaborative model training across multiple medical centers without sharing data, offering significant promise for privacy-preserving AI in healthcare. However, FL models often lack generalization across all participating clients (inside FL) and perform poorly when deployed to unseen clients (outside FL), particularly in heterogeneous domains. Current test-time adaptation methods for outside FL fail to address biases in personalized models toward source distributions, limiting their clinical applications. To tackle these challenges, we propose MSAFed, a generalized multi-stage adaptive FL framework that enhances both inside generalization and outside test-time adaptation. During pretraining, intra-client and inter-client contrastive learning with prototype-aware aggregation produces a generalized global model. An adaptive learning rate strategy further improves inside FL generalization. For unseen clients, source knowledge, including adaptive learning rates and prototypes, is leveraged to dynamically adapt the network architecture during test time. Experiments on three real-world multi-center medical datasets demonstrate the effectiveness of MSAFed, achieving superior performance on both inside and outside FL tasks.

DynFed: Dynamic test-time adaptation for federated learning with adaptive rate networks

The article presents a synthetic set of indicators of the availability of new technologies by aspect digital behaviour at the household’s indicators level. Online activities describe digital behaviour in households by purposes using the internet. Patterns of household behaviour are variable by age, gender, and nationality. The article cites indicators of access to the digital society, pointing to behaviours that confirm the different level of access and use of the Internet in Polish and German households. The juxtaposition of differences in Internet use between countries highlights the different rates of adaptation in the process of digitization in which the availability of digital tool capital becomes a determinant of the use of digital communications and services in modern society. Structure: A comparison of household data for Poland and Germany, drawn from Eurostat, the OECD, and the World Bank for the years 2022–2024. Internet access and online activity reflect digital behaviours in households. Differences in household members’ behaviours indicate distinct patterns of how households are adapting to the digitization process (Gomes, 2024).

This study presents a comprehensive investigation of leukemic B-lymphoblast cell classification using a model-based deep learning pipeline, enhanced with Local Feature Enrichment (LFE), Adaptive Learning Rates (ALR), and Adaptive Channel Recalibration (ACR).Six experimental scenarios were evaluated to analyze the impact of each adaptive mechanism on model performance.Scenarios 1 to 3 employed fixed learning rates without LFE or ACR, yielding moderate performance with accuracy ranging from 92.13% to 93.47%, reflecting limitations in adapting to subtle morphological variations.Scenarios 4 to 6 incorporated adaptive techniques that progressively improved performance, with Scenario 6 achieving peak metrics: accuracy of 99.94%, precision of 99.91%, recall of 99.89%, F1score of 99.90%, and specificity of 99.87%.Comparative evaluation against state-of-the-art methods, including CNNbased ECA modules (91.1%),Vision Transformer (88.2%),Majority Voting Technique (98.5%),EfficientNetB3 (99.31%), and Hybrid Encapsulation EfficientNetV2-S (99.92%), demonstrated that the proposed method outperforms previous approaches, providing superior accuracy, robustness, and stability.External validation using the ALL-IDB2 dataset confirmed that the model maintains high performance across all six scenarios, with accuracy ranging from 90.83% to 97.48%, supporting the generalizability and reliability of the proposed framework.These results confirm that the synergistic integration of adaptive learning strategies and feature enrichment significantly enhances classification performance, offering a reliable framework for automated leukemia detection from blood smear images.The high performance and reliability of this approach can contribute considerably to Public Health by enabling faster, more accurate disease screening at the population level.They may serve as a foundation for future product innovation in diagnostic technologies.

Despite the advancements in federated learning (FL) for privacy-preserving AI, its implementation in medical diagnostics faces persistent challenges. Some of these are intolerable computational loads, communication delays, and inefficiencies in dynamic healthcare environments. Since X-ray models are likely to be imbalanced/heterogeneous, and be prone to adversarial attacks, the likelihood of slow convergence, high complexity, and the negative effects of these with FL models are high. As is highly likely, this will result in performance loss under constraints of such level —most likely with noisy or imbalanced data. Nevertheless, to have a stable FL platform for the actual healthcare utilization, it is essential to overcome these challenges. These challenge are responded to by proposing a new integration of ResNet-50 and privacy preserving techniques for federated medical diagnosis. The latter is a data privacy and security, which is achieved with stochastic noise and homomorphic encryption on top of the superior classification performance of ResNet-50 in processing high dimensional X-ray images. Through effective imbalanced datasets reduction and reduced computational complexity, our method reduces convergence during training with adaptive learning rates and advanced data balancing methods. The model is applied on Python with the use of the TensorFlow and PyTorch libraries, integrating the real-time data using improved communication formula. It is also compared with other approaches and achieve a considerable improvement in accuracy of 99.6%, Precision is 98.8%, Recall is 99.2%, and F1 Score is 99% compared to the 5% improvement of conventional FL models. The model is equally stable and scalable in the dynamic case of a healthcare setting, with guarantees on effective diagnostics and data privacy. The system is a new standard for privacy friendly AI in healthcare diagnosis and gives an effective and scalable solution in X-ray classification task. Now that it’s an open door for future research, there’s further expansion of more complex privacy mechanisms and integration with other AI systems.

Abstract Partitioning scheme in distributed stream processing systems is a critical factor in processing large‐scale real‐time data streams. Existing solutions either suffer from poor partition quality or high overhead in partitioning time, resulting in severe workload imbalance and backpressure problems. This paper proposes an adaptive sampling‐driven workload balancing (ASWB) scheme to solve the above problem. ASWB introduces two core innovations: (1) an adaptive sampling rate algorithm driven by backpressure signals, which dynamically controls the stream sampling rate to avoid prediction bottlenecks and alleviate workload imbalance, and (2) a high‐frequency key prediction module that leverages a variable‐size window to reduce hash collisions and improve frequency estimation accuracy. This design accelerates the identification of high‐frequency keys within the partition operator, thereby enhancing partitioning quality. We implement ASWB on Apache Flink and evaluate it using large‐scale real‐world datasets. Experimental results show that ASWB improves system throughput by up to 55.03% and reduces processing latency compared with state‐of‐the‐art approaches.