Search Results (5,319)

Artificial intelligence (AI) often suffers from high energy consumption and complex deployment in resource-constrained environments, leading to a structural mismatch between capability and deployability. This review takes two representative scenarios—energy-first and performance-first—as the main thread, systematically comparing cloud, edge, and fog/cloudlet/mobile edge computing (MEC)/micro data center (MDC) architectures. Based on a standardized literature search and screening process, three categories of miniaturization strategies are distilled: redundancy compression (e.g., pruning, quantization, and distillation), knowledge transfer (e.g., distillation and parameter-efficient fine-tuning), and hardware–software co-design (e.g., neural architecture search (NAS), compiler-level, and operator-level optimization). The purposes of this review are threefold: (1) to unify the “architecture–strategy–implementation pathway” from a system-level perspective; (2) to establish technology–budget mapping with verifiable quantitative indicators; and (3) to summarize representative pathways for energy- and performance-prioritized scenarios, while highlighting current deficiencies in data disclosure and device-side validation. The findings indicate that, compared with single techniques, cross-layer combined optimization better balances accuracy, latency, and power consumption. Therefore, AI miniaturization should be regarded as a proactive method of structural reconfiguration for large-scale deployment. Future efforts should advance cross-scenario empirical validation and standardized benchmarking, while reinforcing hardware–software co-design. Compared with existing reviews that mostly focus on a single dimension, this review proposes a cross-level framework and design checklist, systematizing scattered optimization methods into reusable engineering pathways. Full article

Densely integrated microarchitectures spanning three-dimensional integrated circuits (3D-ICs), chiplet-based designs, and system-in-package (SiP) assemblies make heat a first-order security concern rather than a mere reliability issue. This review consolidates the landscape of thermal side-channel attacks (TSCAs) on densely integrated microarchitectures: we systematize observation vectors and threat models, clarify core concepts and assumptions, compare the most credible evidence from the past decade, and distill the main classes of defenses across the hardware–software stack. We also explain why hardening against thermal leakage is integral to cyber–physical system (CPS) security and outline the most promising research directions for the field. The strategic relevance of this agenda is reflected in current policy and funding momentum, including initiatives by the United States Department of Homeland Security and the Cybersecurity and Infrastructure Security Agency (DHS/CISA) on operational technology (OT) security, programs by the National Science Foundation (NSF) on CPS, and Canada’s Regional Artificial Intelligence Initiative and Cyber-Physical Resilience Program (RAII, >CAD 35 million), to bridge advanced microelectronics with next-generation cybersecurity. This survey offers a clear, high-level map of the problem space and a focused baseline for future work. Full article

This study evaluates the feasibility of employing rainwater as an alternative feedstock for hydrogen production via electrolysis. While conventional systems typically rely on high-purity water—such as deionized or distilled variants—these can be cost-prohibitive and environmentally intensive. Rainwater, being naturally available and minimally treated, presents a potential sustainable alternative. In this work, a series of comparative experiments was conducted using a proton exchange membrane electrolyzer system operating with both deionized water and rainwater collected from different Austrian locations. The chemical composition of rainwater samples was assessed through inductively coupled plasma, ion chromatography and visual rapid tests to identify impurities and ionic profiles. The electrolyzer’s performance was evaluated under equivalent operating conditions. Results indicate that rainwater, in some cases, yielded comparable or marginally superior efficiency compared to deionized water, attributed to its inherent ionic content. The study also examines the operational risks linked to trace contaminants and explores possible strategies for their mitigation. Full article

There has been a recent shift in the global wine market towards reduced-alcohol wines. Muscadine grapes (Vitis rotundifolia) have become a popular choice in many emerging markets; however, their suitability in reduced-alcohol wine production has not been extensively tested. In this study, methods to reduce ethanol in muscadine wine were compared to determine differences in chemical and sensory attributes and consumer preference. The methods evaluated included full fermentation time with Saccharomyces cerevisiae (control), reduced fermentation time with Saccharomyces cerevisiae (stopped fermentation), fermentation with Saccharomycodes ludwigii yeast (instead of Saccharomyces cerevisiae), and vacuum distillation. The control and distilled wines were fermented for 121 h, Saccharomycodes ludwigii for 45 h, and the stopped fermentation wine for 3 h. Yeast and sugar levels were monitored throughout the fermentation processes using brix measurements and yeast counts. After the fermentation, the color, pH, volatiles, and titratable acidity (TA) were measured. The results showed that Saccharomycodes ludwigii fermented more slowly than Saccharomyces cerevisiae, and that both the stopped fermentation and Saccharomycodes ludwigii wines had lower titratable acidity with a more intense color. The total concentration of volatile compounds for the Saccharomycodes ludwigii wine and the stopped wine were lower than for the distilled and control wines. A consumer panel (n = 92) judged the wine samples on chemical qualities and overall preference. The distilled wine was perceived as more alcoholic compared to the other reduced-alcohol wines. The results showed that the stopped fermentation and Saccharomycodes ludwigii wines were preferred by consumers over the control and vacuum-distilled wines. Full article

This paper introduces a hybrid learning framework that synergistically combines Reinforcement Learning (RL) and Supervised Learning (SL) to train autonomous cyber-defense agents capable of operating effectively in dynamic and adversarial environments. The proposed approach leverages RL for strategic exploration and policy development, while incorporating SL to distill high-reward trajectories into refined policy updates, enhancing sample efficiency, learning stability, and robustness. The framework first targets specialized agent training, where each agent is optimized against a specific adversarial behavior. Subsequently, it is extended to enable the training of a generalized agent that learns to counter multiple, diverse attack strategies through multi-task and curriculum learning techniques. Comprehensive experiments conducted in the CybORG simulation environment demonstrate that the hybrid RL–SL framework consistently outperforms pure RL baselines across both specialized and generalized settings, achieving higher cumulative rewards. Specifically, hybrid-trained agents achieve up to 23% higher cumulative rewards in specialized defense tasks and approximately 18% improvements in generalized defense scenarios compared to RL-only agents. Moreover, incorporating temporal context into the observation space yields a further 4–6% performance gain in policy robustness. Furthermore, we investigate the impact of augmenting the observation space with historical actions and rewards, revealing consistent, albeit incremental, gains in SL-based learning performance. Key contributions of this work include: $\left(i\right)$ a novel hybrid learning paradigm that integrates RL and SL for effective cyber-defense policy learning, $\left(ii\right)$ a scalable extension for training generalized agents across heterogeneous threat models, and $\left(iii\right)$ empirical analysis on the role of temporal context in agent observability and decision-making. Collectively, the results highlight the promise of hybrid learning strategies for building intelligent, resilient, and adaptable cyber-defense systems in evolving threat landscapes. Full article

The main objective of this work is to develop a cost-effective and durable ceramic membrane for water purification. The low-cost ceramic membrane was fabricated using readily available materials, such as clays, aluminum oxide, and calcium carbonate, The membrane was fabricated by uniaxial pressing at different pressures and sintering temperatures, then tested using a scanning electron microscope (SEM) and XRD. The porosity of the resulting membrane was 38.7%, and the contact angle was 65° indicating good hydrophilicity for filtration applications. The main composition was 70% clay, 25% CaCO₃, and 5% alumina. The removal % for methylene blue was tested at varying concentrations, achieving up to 99% removal, an initial flux of 496.8 L m⁻² h⁻¹, and an average pore size of 2 µm. Furthermore, the research explores the effect of backwashing cycles and techniques on the membrane long-term performance. The results indicated that washing the membrane for four cycles to cleanness has achieved an improved efficiency of the membrane and % dye rejection. Back washing was achieved using no chemicals; only distilled water and drying were used. A preliminary costs assessment of the production for affordable membrane resulted in a value of 170 USD/m². The findings demonstrate that optimizing backwashing cycles is essential for prolonging the membrane lifespan and lowering operation costs. Full article

The present research describes the chemical composition and the enantiomeric profile of a spicy green aroma essential oil, distilled from the dry leaves of Baccharis sinuata Kunth (Asteraceae). The distillation yield was as high as 3.0% by weight. The chemical analysis was conducted on two columns, coated with stationary phases of different polarity (5% phenyl—95% methyl polysiloxane, expressed by weight, and 100% polyethylene glycol). Major components (≥2.0% as an average value between the two columns) were as follows: β-pinene (4.9%), limonene (39.0%), (E)-β-caryophyllene (2.0%), bicyclogermacrene (2.7%), γ-cadinene (4.0%), δ-cadinene (7.3%), β-eudesmol (2.0%), α-eudesmol (3.0%), and α-cadinol (2.0%). For the enantioselective analysis, 10 enantiomeric pairs were investigated, using two capillary columns coated with different chiral selectors. As a result, (1R,5R)-(−)-α-thujene, (1S,5S)-(−)-α-pinene, and (1R,2S,6S,7S,8S)-(−)-α-copaene were enantiomerically pure, whereas (R)-(+)-limonene presented a 90.0% enantiomeric excess. All the other analysed chiral compounds were scalemic mixtures. The high distillation yield, its aroma, and the bibliographic bioactivity profile make this essential oil potentially interesting from a commercial point of view. To the best of the authors’ knowledge, this is the first description of an essential oil distilled from leaves of B. sinuata. Full article

With the increasing demand to scale the chip industry, attention is turning to the vital role that phosphanes and silanes play in semiconductor manufacturing processes such as chemical vapor deposition, plasma etching, and impurity doping. High-performance semiconductors often require a supply of ultra-pure gaseous phosphine (≥99.999%) to ensure the formation of defect-free thin-film structures with high integrity and strong functionality. In recent years, research on high-purity PH₃ synthesis methods has mainly focused on two pathways: the acidic route with fewer side reactions, high by-product economics, and higher exergy of high-purity PH₃, and the alkaline alternative with greater potential for practical application through lower reaction temperatures and a simpler reaction process. This paper presents the first comparative study and analysis on the preparation of ultra-high-purity PH₃ and its process energy consumption. Using Aspen and its related software, the energy consumption and cost issues are discussed, and the process heat exchange network is established and optimised. By combining Aspen Plus V14 with MATLAB 2023, an artificial neural network (ANN) prediction model is established, and the parameters of the distillation section equipment are optimised through the NSGA-II model to solve problems such as low product yield and large equipment exergy loss. After optimisation, it can be found that in terms of energy consumption and cost indicators, the acidic process has greater advantages in large-scale production of high-purity PH₃. The total energy consumption of the acidic process is 1.6 × 10⁸ kJ/h, which is only one-third that of the alkaline process, while the cost of the heat exchange equipment is approximately three-quarters that of the alkaline process. Through dual-objective optimisation, the exergy loss of the acidic distillation part can be reduced by 1714.1 kW, and the economic cost can be reduced by USD 3673. Therefore, from the perspective of energy usage and equipment manufacturing, the comprehensive analysis of the acidic process has more advantages than that of the alkaline process. Full article

The durations of wars fought between 1480 and 1941 A.D. were found to be well represented by random numbers chosen from a single-event Poisson distribution with a half-life of (1.25 ± 0.1) years. This result complements the work of L.F. Richardson who found that the frequency of outbreaks of wars can be described as a Poisson process. This result suggests that a quick return on investment requires a distillation of the many stressors of the day, each one of which has a small probability of being included in a convincing well-orchestrated simple call-to-arms. The half-life is a measure of how this call wanes with time. Full article

Rosa bracteata J.C.Wendl. is a thorny, clump-forming or trailing perennial evergreen shrub native to China. The current analysis was designed to explore the chemical constituents and determine the in vitro antimicrobial, cytotoxic, and antioxidant properties of the essential oils (EOs) of the stems, leaves, and flowers of Rosa bracteata for the first time. The chemical composition of the essential oils obtained through hydro-distillation was characterized by means of gas chromatography–mass spectrometry (GC–MS) and gas chromatography with a flame ionization detector (GC–FID). Thirty-seven, thirty-six, and forty-two constituents were identified from leaf oil (LEO), stem oil (SEO), and flower oil (FEO), representing 96.3%, 95.9%, and 97.4% of the total oil constituents, respectively. The LEO was mainly composed of 1-pentadecene, α-cadinol, and hexadecanoic acid. However, the main identified components of SEO were (E)-nerolidol, phytol, and benzyl benzoate, and the main components of the flower oil were ethyl octanoate, octanoic acid, and α-cadinol. All of the EOs exhibited antibacterial activities against both Gram-positive and Gram-negative bacteria with MIC values ranging from 40.00 to 640.00 μg/mL. In addition, the checkerboard method demonstrates potent synergistic effects of Rosa bracteata EOs when combined with commercial antibiotics (chloramphenicol and streptomycin). In the MTT test, SEO (IC₅₀: 37.91 ± 2.10 to 51.15 ± 6.42 μg/mL) showed stronger cytotoxic activity against four cancer cell lines (MCF-7, A549, HepG2, and HCT-116) during the incubation time of 48 h in comparison to the EOs isolated from the other plant parts. Overall, these findings reveal the chemical composition and significant bioactivity of R. bracteata EOs for the first time, suggesting their potential as promising natural agents for therapeutic applications, especially in combination therapies to combat antibiotic resistance. Full article

Since H₂O and Ethylene Glycol Monomethyl Ether (EM) form a minimum-boiling azeotrope, 1-pentanol, 1-hexanol, and 1-heptanol are selected as entrainers to separate the azeotropic mixture (H₂O/EM) using azeotropic distillation. The binary vapor liquid equilibrium (VLE) data were determined at 101.3 kPa, including H₂O/EM, EM/1-pentanol, EM/1-hexanol, EM/1-heptanol, H₂O/1-pentanol, H₂O/1-hexanol and H₂O/1-heptanol. Meanwhile, the Herington area test was used to validate the thermodynamic consistency of the experimental binary data. The VLE data for the experimental binary system were analyzed using the NRTL, UNIQUAC, and Wilson activity coefficient models, showing excellent agreement between predictions and measurements. Finally, molecular simulations were employed to calculate interaction energies between components, providing insights into the VLE behavior. The azeotropic distillation process was simulated using Aspen Plus to evaluate the separation performance and determine the optimal operating parameters. Therefore, this study provides guidance and a foundational basis for the separation of H₂O/EM systems at 101.3 kPa. Full article

Humans learn new tasks without forgetting, but neural networks suffer from catastrophic forgetting when trained sequentially. Dynamic expandable networks attempt to address this by assigning each task its own feature extractor and freezing previous ones to preserve past knowledge. While effective for retaining old tasks, this design leads to rapid parameter growth, and frozen extractors never adapt to future data, often producing irrelevant features that degrade later performance. To overcome these limitations, we propose Task-Sharing Distillation (TSD), which reduces the number of extractors by allowing multiple tasks to share one extractor and consolidating them through distillation. We study two strategies: (1) grouped rolling consolidation, which groups consecutive tasks and consolidates them into a shared extractor, and (2) a fixed-size pooling with similarity-based consolidation, where new tasks are merged into the most compatible extractor in a fixed pool according to prototype similarity. Experiments on the CIFAR-100 and ImageNet-100 datasets show that TSD maintains strong performance across tasks, demonstrating that careful feature sharing is more effective than simply adding more extractors. On ImageNet-100, our method achieves 2.5% higher average accuracy than DER while using fewer feature extractors. Full article

This study aimed to evaluate the impact of different adhesive solutions on the surface roughness and hardness of denture base materials. Twenty specimens (20 × 20 × 5 mm) were produced for each material group: heat-cured ProBase Hot^®, 3D-printed NextDent Denture 3D+^®, and PMMA-milled Exaktus^®. They were then divided into five solution subgroups (n = 4): control (T0), distilled water, Corega PowerMax^®, Elgydium Fix^®, and Kukident Pro Ultimate^®. Specimens were immersed in the solution at 37 °C daily for 28 days, simulating continuous use. Profilometry and Shore D hardness tests were performed at baseline and after 28 days of the immersion protocol. Data analysis was done using IBM SPSS Statistics 30.0, considering a confidence level of 0.05. At baseline, the materials differed in surface roughness and Shore D hardness, with the 3D-printed group showing the highest median values for the Rz parameter (p = 0.023) and the lowest for hardness (p = 0.023). Elgydium Fix had a significant effect on the heat-cured resin, with increased Rz and decreased hardness. Kukident caused higher roughness and lower hardness in the 3D-printed and milled resins (not significant). Corega showed minor effects in all tested materials. In conclusion, the denture base material and the adhesive formulation influence the physical and mechanical properties of denture base resins. Full article

This study valorizes onion peel, an agro-industrial by-product rich in phenolic compounds and structural carbohydrates, for the development of cassava starch-based biodegradable films. The films were prepared using the solution casting method; a cassava starch matrix was mixed with a 2.5% glycerol solution and heated to 85 °C for 30 min. A separate solution of onion peel powder (OPP) in distilled water was prepared at 25 °C. The two solutions were then combined and stirred for an additional 2 min before 25 mL of the final mixture was cast to form the films. Onion peel powder (OPP) incorporation produced darker and more opaque films, suitable for packaging light-sensitive foods. Film thickness increased with OPP content (0.138–0.218 mm), while moisture content (19.2–32.6%) and solubility (24.0–25.2%) decreased. Conversely, water vapor permeability (WVP) significantly increased (1.69 × 10⁻⁹–2.77 × 10⁻⁹ g·m⁻¹·s⁻¹·Pa⁻¹; p < 0.0001), reflecting the hydrophilic nature of OPP. Thermal analysis (TGA/DSC) indicated stability up to 245 °C, supporting applications as food coatings. Morphological analysis (SEM) revealed OPP microparticles embedded in the starch matrix, with FTIR and XRD suggesting electrostatic and hydrogen–bond interactions. Mechanically, tensile strength improved (up to 2.71 MPa) while elongation decreased (14.1%), indicating stronger but less flexible films. Biodegradability assays showed slightly reduced degradation (29.0–31.8%) compared with the control (38.4%), likely due to antimicrobial phenolics inhibiting soil microbiota. Overall, OPP and cassava starch represent low-cost, abundant raw materials for the formulation of functional biopolymer films with potential in sustainable food packaging. Full article

In the face of ever-evolving cyber threats, modern intrusion detection systems (IDS) must achieve long-term adaptability without sacrificing performance on previously encountered attacks. Traditional IDS approaches often rely on static training assumptions, making them prone to forgetting old patterns, underperforming in label-scarce conditions, and struggling with imbalanced class distributions as new attacks emerge. To overcome these limitations, we present a continual learning framework tailored for adaptive intrusion detection. Unlike prior methods, our approach is designed to operate under real-world network conditions characterized by high-dimensional, sparse traffic data and task-agnostic learning sequences. The framework combines three core components: a clustering-based memory strategy that selectively retains informative historical samples using DP-Means; multi-level knowledge distillation that aligns current and previous model states at output and intermediate feature levels; and a meta-learning-driven class reweighting mechanism that dynamically adjusts to shifting attack distributions. Empirical evaluations on benchmark intrusion detection datasets demonstrate the framework’s ability to maintain high detection accuracy while effectively mitigating forgetting. Notably, it delivers reliable performance in continually changing environments where the availability of labeled data is limited, making it well-suited for real-world cybersecurity systems. Full article