Hong Kong University of Science and Technology
Hong Kong University of Science and Technology Institutional RepositoryNot a member yet
162821 research outputs found
Sort by
MLFD: a novel meta-learning method with fourier transform data augmentation for domain generalization
No full textAs Machine Learning (ML) and Artificial Intelligence (AI) progress rapidly, the issue of ML model generalization has emerged as a critical concern for academics and practitioners alike. In practical scenarios, it is essential for models to sustain high performance when encountering varied and novel data distributions. Nevertheless, current domain generalization techniques have their shortcomings in tackling this challenge. The objective of this paper is to introduce a novel Meta-Learning approach, incorporating Fourier transform-based Data augmentation, called MLFD, for the purpose of domain generalization. Utilizing both data augmentation and a meta-learning architecture, this proposed technique empowers models to extend their generalization to multiple unseen target domains using just a single training domain. In contrast to other domain generalization methods, the method presented in this paper achieves comparable accuracy on the Digits-DG datasets, and demonstrates substantial improvements in terms of reducing model training time.</p
The role of particle shape in stress transmission and stiffness in gap-graded soils
No full textThis study investigates how particle shape governs the mechanical behavior of gap-graded soils, with a focus on stress transmission, small-strain stiffness, and the efficacy of a refined state variable framework. Based on systematic DEM simulations with non-spherical particles, we demonstrate that non-spherical coarse particles significantly enhance the engagement of finer particles in stress transmission, particularly when the fine content is between 15% and 25%. This effect stems from shape-induced changes in packing structure, which alter the void ratio and coordination number. In contrast, the effect of particle shape on small-strain stiffness, while measurable, is quantitatively limited. This is because the stiffness contribution from finer particles remains minor, even when their role in stress transmission is amplified. The refined state variable framework, which account for this enhanced role of fines, was validated and effectively captured the small-strain characteristics of gap-graded soils across all particle shapes. Overall, the observed mechanical trends remain consistent with those from studies using spherical particles, indicating that the underlying mechanisms are robust. These insights improve the understanding of coupled shape–size effects and provide enhanced modeling tools for geotechnical applications such as foundation design and material optimization.</p
Mechanically gated OSCA/TMEM63 ion channels: from physiological function to structural basis
No full textThe OSCA/TMEM63 family constitutes the largest identified group of mechanically gated ion channels, having pivotal roles in cellular mechanotransduction. OSCAs are osmosensitive and mechanically gated ion channels in plants, while their animal homologs TMEM63s similarly respond to mechanical stimuli and osmotic stress, characterized by smaller conductance and higher mechanoactivation thresholds. Structural studies suggest a conserved "force-from-lipid" gating mechanism for the OSCA/TMEM63 family, in which membrane tension or lipid bilayer curvature directly induces conformational changes and ion conductance. This review summarizes the physiological and pathological roles of OSCA/TMEM63 proteins, highlighting potential molecular mechanisms for their roles in mechanotransduction. Based on the structural homology and functional properties in biological membranes, we proposed a mechanically gated ion channel superfamily comprising OSCA/TMEM63 and TMC proteins.</p
Self-supported partially crystallized nanoporous metallic glass for ultra-stable and efficient electrocatalytic hydrogen evolution
No full textMetallic glasses (MGs) often suffer from sluggish hydrogen evolution reaction5 (HER) kinetics in neutral and alkaline media, with their catalytic performance predominantly confined to acidic environments. Herein, we reported a novel thermoplastic forming technique to fabricate a self-supported partially crystallized nanoporous Pt56.2Ni5.2Cu16.8P21.8 metallic glass (C-NPMG). The C-NPMG catalyst delivers ultralow overpotentials of 18.0 mV (0.5 M H2SO4), 42.2 mV (1 M KOH), and 88.0 mV (1 M phosphate-buffered saline (PBS)) at a current density of 10 mA cm−2, outperforming most state-of-the-art non-noble MGs and Pt-based benchmarks across all pH conditions. Notably, it maintains negligible performance decay for over 1000 h in alkaline electrolytes, showcasing superior stability. Experimental and computational analyses reveal that the enhanced HER activity arises from three synergistic effects: (1) the high-specific-surface-area nanoporous architecture that maximizes active site exposure; (2) the formation of crystallite-amorphous interfaces during partial crystallization, which lowers the energy barrier for H2 desorption; (3) the hierarchical super-hydrophilic and super-hydrophobic wettability of the C-NPMG, which optimizes mass transport and prevents electrolyte-induced corrosion. This work establishes a novel design paradigm for developing high-performance, pH-universal HER electrocatalysts by integrating structural nano-engineering and crystallite-amorphous phase synergy in metallic glass systems to overcome the trade-offs between performance and stability in electrochemical water splitting. (Figure presented.
Multifunctional hydrogel platform encapsulating lotus root-derived extracellular vesicles stimulate osteogenesis-angiogenesis coupling for accelerated bone regeneration
No full textBone defects due to aging, trauma, and diseases pose a major clinical challenge, causing both physical and psychological distress for patients. Current clinical treatments including autologous or allogeneic bone grafting are limited by donor scarcity. Bone regeneration involves multi-system crosstalk, therefore hydrogel-integrated therapeutic platform strategies with multifunctional and graded sequential degradation are particularly promising. Plant-derived extracellular vesicles (PEVs) have attracted increasing attention in tissue regeneration due to their various bioactivities, easy accessibility, and low immunogenicity, yet suffer from the drawbacks of rapid clearance in vivo. In this study, we extracted extracellular vesicles from lotus roots (LREVs) for the first time and developed a multifunctional hydrogel platform (GelMA/LREVs) incorporating them. The initial in vitro experiments demonstrated that this platform effectively scavenged excessive reactive oxygen species (ROS) in BMSCs, upregulated osteogenesis-angiogenesis coupling, and activated the TGF-β/Smad signaling pathway to promote osteogenic differentiation. Further in vivo experiments confirmed that the platform significantly accelerated bone regeneration and maturation by enhancing angiogenesis, stimulating osteogenic differentiation, and facilitating collagen maturation and mineralization. Given these excellent properties, this platform is anticipated to offer a promising strategy for the clinical treatment of critical bone defects.</p
Pulsar polarization array limits on ultralight axionlike dark matter
No full textWe conduct the first-ever pulsar polarization array (PPA) analysis to detect the ultralight axionlike dark matter (ALDM) using the polarization data of 22 millisecond pulsars from the third data release of Parkes pulsar timing array. As one of the major dark matter candidates, the ultralight ALDM exhibits a pronounced wave nature on astronomical scales and offers a promising solution to small-scale structure issues within local galaxies. While the linearly polarized pulsar light travels through the ALDM galactic halo, its position angle (PA) can be subject to an oscillation induced by the ALDM Chern-Simons coupling with electromagnetic field. The PPA is thus especially suited for detecting the ultralight ALDM by correlating polarization data across the arrayed pulsars. To accomplish this task, we develop an advanced Bayesian analysis framework that allows us to construct pulsar PA residual time series, model noise contributions properly and search for pulsar cross-correlations. We find that for an ALDM density of ρ0=0.4 GeV/cm3, the Parkes PPA offers the best global limits on the ALDM Chern-Simons coupling, namely ≲10-13.5-10-12.2 GeV-1, for the mass range of 10-22-10-21 eV. The crucial role of pulsar cross-correlation in recognizing the nature of the derived limits is also highlighted.</p
Tailored charging protocol for densified lithium deposition and stable initially anode-free lithium metal pouch cells
No full textThe low mechanical strength and insufficient lithium-ion conductivity of solid-electrolyte interphase will incur inhomogeneous lithium deposition and eventually cause battery failure, hindering practical application of lithium metal batteries. Combining multiple characterization techniques and COMSOL simulation, here we report that the commonly-adopted constant current charge protocol would result in formation of highly porous lithium metal negative electrode, leading to severe parasitic reactions. To ameliorate this problem, we propose a middle peak current charge protocol with a peak current in the middle of charging step, which could guide the nucleation and growth of densified lithium deposition with the derived solid-electrolyte interphase possessing significantly improved mechanical strength and Li+ ion conductivity. Therefore, coin-type initially anode-free lithium metal batteries could be stably cycled for 80 cycles (with 3 hours of charging and state-of-charge of ~70%). A 1.5 Ah initially anode-free Li metal pouch cell which delivers a specific energy of 400 Wh kg-1 at 225 mA and with a capacity retention of 80% for 298 cycles (with 3 hours of charging and state-of-charge of 80%) is also demonstrated.</p
The effects of nutrition support on behavioral outcomes and labor productivity
No full textWhile many of the world's poor consume inadequate calories with low nutritional value, there has been little work on how this may shape their behaviors and productivity. Using lab-in-the-field and field experiments in the context of a floriculture plant in Ethiopia, this study investigates the effects of a nutrition support program on behavioral outcomes—stress, prosociality, cooperation, and attention—and productivity. We find that nutrition support relieves stress and decreases prosociality, exhibiting a pattern of hedonic adaptation over time. We do not find evidence for improvements in labor productivity.</p
Microphysical analysis on the impacts of reduced aerosol concentrations on tropical cyclone precipitation in South China Area
No full textPrevious studies have discovered that an increase in aerosol concentration could affect the development of precipitation and tropical cyclones. However, with increasing efforts to mitigate climate change, the amount of anthropogenic aerosols in the atmosphere is expected to decrease in the future. This study aims at understanding how such a reduction in aerosol number concentration could affect the precipitation pattern of tropical cyclones through case studies of Typhoon Haikui and Koinu. The Thompson aerosol-aware microphysics scheme in the Weather Research and Forecast (WRF) model is used, and the water-friendly aerosol number concentration is reduced by a factor of 100. For both tropical cyclones, a reduction in aerosol results in consistent expansion of precipitation area due to the enhanced warm rain process. However, the total precipitation of Koinu decreases while that of Haikui increases, depending on whether warm-rain or ice-phase processes dominate and whether upper-level convection is altered.</p
