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
VOCs-driven ozone extremes during dry and wet heatwaves in the Jiangsu–Shandong–Henan–Anhui Boundary: Integrating meteorological forcings and SHAP interpretation
No full textClimate change is driving more frequent and severe dry and wet heatwaves, yet a clear picture of how each type influences ozone (O3) production in areas characterized by intricate industry and geography are still lacking. This study examines O3 formation during heatwaves, focusing on interactions among meteorological factors, atmospheric chemistry, and pollutant emissions in a unique industrial area at the junction of Jiangsu, Shandong, Henan, and Anhui provinces of northern China in summer 2022 and 2023. This study integrates hourly data of temperature, relative humidity (RH), solar radiation (SF), 115 VOCs, and other atmospheric pollutants, and quantifies the contribution of each factor using machine learning models combined with SHAP. Results show that SF is the main driver influencing O3, contributing 13.7 μg/m3 during dry heatwaves and 5.0 μg/m3 during wet heatwaves. RH and atmospheric diffusion conditions are distinct between the dry and wet heatwaves. PMF indicates that industrial emissions dominate O3 formation during dry heatwaves while biogenic VOCs dominate during wet heatwaves. For VOCs, during dry heatwaves the SHAP values for styrene, propene and isoprene were 9.1, 4.4 and 3.8 μg/m3, respectively, significantly affecting O3 formation; In wet heatwaves, styrene, propene and acetaldehyde dominate, with SHAP values of 6.7, 4.3 and 2.5 μg/m3 respectively. Diurnal analysis indicates that while styrene and propene boost O3 during daytime (9:00–17:00), their effects reverse in the early morning (6:00–8:00). In contrast, isoprene contributes positively (2.85 μg/m3) during dry heatwaves and negatively (−2.92 μg/m3) during wet ones. Overall, the study offers an efficient framework for understanding O3 formation in extreme weather and informs targeted pollution control strategies.</p
Towards Scientific Machine Learning for Granular Material Simulations: Challenges and Opportunities
No full textMicro-scale mechanisms, such as inter-particle and particle-fluid interactions, govern the behaviour of granular systems. While particle-scale simulations provide detailed insights into these interactions, their computational cost is often prohibitive. At a recent Lorentz Center Workshop on “Machine Learning for Discrete Granular Media”, researchers explored how machine learning approaches can aid the development of constitutive laws and efficient data-driven surrogates for granular materials while also addressing uncertainty quantification. Attended by researchers from both the granular materials (GM) and machine learning (ML) communities, the workshop brought the ML community up to date with GM challenges. This position paper emerged from the workshop discussions. In this position paper, we define granular materials and identify seven key challenges that characterise their distinctive behaviour across various scales and regimes–ranging from gas-like to fluid-like and solid-like. Addressing these challenges is essential for developing robust and efficient models for the digital twinning of granular systems in various industrial applications. To showcase the potential of ML to the GM community, we present classical and emerging machine/deep learning techniques that have been, or could be, applied to granular materials. We reviewed sequence-based learning models for path-dependent constitutive behaviour, followed by encoder-decoder type models for representing high-dimensional data in reduced spaces. We then explore graph neural networks and recent advances in neural operator learning. The latter captures the emerging field evolution of interacting particles via efficient latent space representation. Lastly, we discuss model-order reduction and probabilistic learning techniques for high-dimensional parameterised systems, both of which are crucial for quantifying and incorporating uncertainties arising from physics-based and data-driven models. We present a typical workflow aimed at unifying data structures and modelling pipelines and guiding readers through the selection, training, and deployment of ML surrogates for granular material simulations. Finally, we illustrate the workflow’s practical use with two representative examples, focusing on granular materials in solid-like and fluid-like regimes.</p
Evaluation of high-resolution WRF simulation in urban areas — Effect of different physics schemes on simulation performance in the Rhine-Main-Neckar area
No full textQuantifying and minimizing atmospheric transport errors is key to improve meteorological modeling and to better estimate urban greenhouse gas (GHG) and air pollution emissions from measurements. The Weather Research and Forecasting Model (WRF) model has been used to simulate urban atmospheric transport in many cities globally and there exist various possible configurations especially concerning choice of physics schemes, which influence the quality of the atmospheric simulation. Here, we conduct a comprehensive evaluation of WRF on 1 km resolution for a polycentric European metropolitan area, namely the Rhine-Main-Neckar area by varying Land Surface Model (LSM), Surface Layer Model (SLM), Planetary Boundary Layer (PBL) and urban parametrization scheme configurations. We compare four month-long simulations to 2 m temperature, 10 m wind velocity and wind direction measured at 19 stations operated by the German Weather Service and to PBL height derived from radiosonde data at two locations. By showing kernel density functions in a Taylor diagram, we show the average performance of the schemes as well as the spread across different stations. We find that while the 2 m temperature and PBL height performance are most sensitive to choice of urban parametrization scheme, 10 m wind velocity and direction are most sensitive to choice of PBL scheme. Good overall performance was achieved using the Single-Layer Urban Canopy Model (SLUCM), Mellor-Yamada-Janjic (MYJ), Noah-Multiparametrization Land Surface Model (Noah-MP) and Monin-Obukhov (Janjic) (MO) schemes. While the ensemble spread is larger in winter than in summer, the choice of optimal scheme does not depend strongly on the season.</p
SegMamba-V2: Long-range Sequential Modeling Mamba For General 3D Medical Image Segmentation
No full textThe Transformer architecture has demonstrated remarkable results in 3D medical image segmentation due to its capability of modeling global relationships. However, it poses a significant computational burden when processing high-dimensional medical images. Mamba, as a State Space Model (SSM), has recently emerged as a notable approach for modeling long-range dependencies in sequential data. Although a substantial amount of Mamba-based research has focused on natural language and 2D image processing, few studies explore the capability of Mamba on 3D medical images. In this paper, we propose SegMamba-V2, a novel 3D medical image segmentation model, to effectively capture long-range dependencies within whole-volume features at each scale. To achieve this goal, we first devise a hierarchical scale downsampling strategy to enhance the receptive field and mitigate information loss during downsampling. Furthermore, we design a novel tri-orientated spatial Mamba block that extends the global dependency modeling process from one plane to three orthogonal planes to improve feature representation capability. Moreover, we collect and annotate a large-scale dataset (named CRC-2000) with fine-grained categories to facilitate benchmarking evaluation in 3D colorectal cancer (CRC) segmentation. We evaluate the effectiveness of our SegMamba-V2 on CRC-2000 and three other large-scale 3D medical image segmentation datasets, covering various modalities, organs, and segmentation targets. Experimental results demonstrate that our Segmamba-V2 outperforms state-of-the-art methods by a significant margin, which indicates the universality and effectiveness of the proposed model on 3D medical image segmentation tasks. The code for SegMamba-V2 is publicly available at: https://github.com/ge-xing/SegMamba-V2
SegMIC: A universal model for medical image segmentation through in-context learning
No full textMedical images encompass a wide array of modalities and anatomical structures, requiring numerous segmentation tasks. However, current methods tend to be highly specialized, struggling to generalize to Out-of-Distribution tasks without retraining. Developing a universal segmentation model is thus a valuable yet challenging goal. In this work, we introduce SegMIC, a novel learning paradigm that addresses these challenges through In-Context Learning. The primary aim of SegMIC is to perform segmentation across arbitrary anatomies by harnessing contextual information derived from a single reference image and its corresponding annotation. Building on this concept, SegMIC employs a tailored joint mask on the image-annotation pairs of both the query and the reference, compelling the model to perform tasks conditioned on visible contextual annotation patches. Thus the inference process is streamlined, requiring only a single reference pair to specify the desired task. To train our SegMIC, we compiled an unprecedented large-scale segmentation benchmark (UniMedDB), comprising totaling 49k images and 82k annotations spanning over 14 modalities and dozens of anatomical structures. Comprehensive experiments demonstrate that SegMIC adeptly utilizes context to handle diverse segmentation tasks across various modalities and anatomies, achieving superior Dice and IoU scores of 0.919 and 0.870 on in-distribution tasks while outperforming the best competitors on out-of-distribution tasks by margins of 0.111 in Dice and 0.146 in IoU. Our code and datasets are available at https://github.com/JWZhao-uestc/SegMIC.</p
SQLLM: A Secure and Quantized Framework for Large Language Models in 5G Private Network Operations
No full textWith the development of 5G technology, the demand for large language models (LLMs) in private network operation and maintenance is growing. These models enhance the intelligence and efficiency of network management through deep learning techniques. However, the application of LLMs for 5G private network operation and maintenance faces the dual challenges of data security and resource limitations. To address these challenges, we propose SQLLM, which is an integrated framework dedicated to detecting abnormal user query-based network attacks and performing high-quality compression on large language detection models for network attacks. Specifically, leveraging the powerful representational capacity of large language models, we utilize the LoRA fine-tuning technique to identify normal questions and three types of attack questions, thus avoiding behaviors that attempt to disrupt or obtain sensitive information. Moreover, this research innovatively adopts the static quantization technique to compress LLMs for private network operation and maintenance. Unlike traditional static quantization that uses fixed parameters and uniform granularity for activations, our optimized strategy introduces a dynamic smoothing factor α to transfer outlier variance from activations to weights and adopts a hybrid per-tensor/per-token granularity tailored to 5G O&M data, thus avoiding the challenge of the sharp decline in the inference ability of the quantized LLM caused by outliers, enabling it to adapt to resource-constrained and multi-user concurrent usage environments. Finally, we validate four types of questions and evaluate the performance of the quantized model on key performance indicators such as accuracy, VRAM consumption, and inference time. The experimental results demonstrate that the SQLLM model achieves excellent performance on all test indicators. This achievement proves the security and feasibility of the efficient deployment of LLMs for private network operation and maintenance in the 5G private network environment with network attacks and resource limitations.</p
Transmissive Multilayer Geometric Phase Gratings Using Water-Soluble Alignment Material
No full textMultilayer liquid crystal devices can offer enhanced optical functionalities for augmented reality and photonic applications, but fabrication remains severely limited by solvent incompatibility between photoalignment materials and underlying polymerized layers. Conventional photoalignment agents use aggressive solvents like N,N-dimethylformamide that damage polymerized substrates, necessitating protective interlayers. This study demonstrates a water-soluble photoalignment approach using AbA-2522 that eliminates these fabrication barriers. The water-soluble alignment material enables direct multilayer processing without layer damage while maintaining alignment quality equivalent to conventional materials. We successfully fabricate compact transmissive devices integrating liquid crystal polarization gratings with quarter-wave plates, achieving a first-order diffraction efficiency of 65.4% for 9 μm period gratings for linearly polarized incident light (λ = 457 nm). The multilayer structure exhibits highly selective polarization-dependent diffraction with efficiency ratios exceeding 10:1 between preferred and suppressed orders, eliminating external polarization control elements. Polarized optical microscopy confirms excellent alignment uniformity, while the fabrication process offers environmental benefits and reduced complexity. This approach establishes a practical pathway for advanced multilayer photonic devices critical for next-generation augmented reality systems and photonic integration, addressing fundamental challenges that have limited multilayer liquid crystal device development.</p
Navigating Belonging in Hong Kong: Transcultural Capital, the Authenticity Dilemma and Collaborative Arts-based Research
No full textThis paper presents a study of what it means to belong, for young people from South Asian backgrounds in Hong Kong. The work draws from a collaborative visual and linguistic ethnographic arts-based research project called Navigating Belonging, which responded to ongoing and historical oppression of minoritised ethnic groups in Hong Kong. Through narrative generated in interaction and through participatory photography, participants explored their understandings of belonging. In this paper we pay close attention to the narratives of belonging that emerged during the project’s creative practice workshops, alongside the writing, photography, and digital storytelling that were the creative outputs of the project. We focus on the talk of two participants, Rani and Amir, undergraduate students at a university in Hong Kong. In our analysis we invoke the concept of transculturality. We describe and explain how–through their narratives–our participants deploy their transcultural capital as they address questions of the authenticity of their belonging. In our conclusions we highlight the transformative power of arts-based ethnographic research to engender critical reflection that challenges essentialist notions of identity and belonging, for minoritised youth of South Asian descent in Hong Kong. Finally, we suggest implications of our findings for policy development for this population.</p
Nanotechnology-driven biomaterials for chronic liver diseases: Stage-specific strategies for advanced theranostics
No full textChronic liver diseases (CLDs), encompassing a spectrum from steatosis and inflammation to fibrosis, cirrhosis, represent a major global health burden, causing approximately 2 million deaths annually [1]. The management of CLDs is significantly hampered by the limitations of conventional approaches, including non-targeted drug delivery, systemic toxicity, and inadequate diagnostic sensitivity for early-stage lesions. Nanotechnology-driven biomaterial platforms have emerged as pioneering solutions to these challenges, enabling precise theranostic strategies tailored to the distinct pathophysiology of each disease stage. This review systematically elaborates on these advancements by aligning with the natural progression of CLDs [non-alcoholic fatty liver disease (NAFLD), non-alcoholic steatohepatitis (NASH), hepatitis B, liver fibrosis, and cirrhosis]. We detail how engineered platforms enhance therapeutic efficacy by achieving superior hepatic accumulation, controlled drug release, and improved metabolic, antiviral, and antifibrotic effects. Concurrently, we explore their role in diagnostics, where nanotechnology-enhanced imaging agents and nanosensors provide unprecedented sensitivity for early detection and accurate staging. By structuring the discussion around the evolving clinical needs from NAFLD and hepatitis to advanced fibrosis and cirrhosis, this review offers a stage-specific roadmap of biomaterial design principles. It aims to provide a foundational theory and forward-looking perspectives for developing next-generation, precision medicine solutions for CLDs, ultimately bridging the gap between benchtop innovation and clinical translation. Statement of Significance This review establishes a stage-specific design paradigm that bridges the gap between biomaterial innovation and the clinical continuum of chronic liver diseases (CLDs). Its significance lies in aligning cutting-edge biomaterial strategies from targeted, stimuli-responsive nanotherapeutics to engineered exosomes and gene delivery systems with the distinct pathophysiological features of each disease stage. This approach moves beyond cataloging materials to critically evaluating their translational feasibility. We analyze how rational material design addresses specific clinical bottlenecks, such as improving drug bioavailability to diseased tissue or enabling sensitive, non-invasive diagnostics for early detection. By providing this clinically focused roadmap, this review aims to accelerate the development of personalized therapies and reshape the theranostic landscape, striving to improve therapeutic outcomes of CLDs.</p
Mental health, psychological wellbeing, and coping with stress by Ukrainian war refugees staying in Poland
No full textBackground: The ongoing Russian-Ukrainian conflict has led to the large-scale displacement of over one million Ukrainian citizens, predominantly women and children, seeking protection in Poland. This necessitates a rigorous investigation into the factors associated with psychological conditions and adaptive coping mechanisms employed by this vulnerable population. Methods: This quantitative, online survey study comprised a sample of N = 290 adult participants (91.7% female, mean age 43.6 years). Due to the small number of male participants (n = 24, 8.3%), gender was not included as a variable in the analyses. Participants were recruited through social workers and psychologists working in Collective Accommodation Centers, and Integration Centers in Poland for Ukrainian refugees. Depression, Anxiety, and Stress symptoms were assessed using the DASS-21, psychological wellbeing with the PERMA-Profiler, and coping with stress strategies using the Brief-COPE. Results: Analysis revealed significantly elevated psychological distress among Ukrainian refugees in Poland. Multivariate regression models identified independent predictors of mental health, wellbeing and coping with stress. Older age, partnered status, and skills-matched employment are key predictors of depression. Higher education and partnered status are negative predictors of anxiety, and older age is a negative predictor of stress. Skills-matched employment emerged as a predictor of wellbeing, though only 23.5% of this highly educated sample held such positions. Access to information, receipt of psychological assistance, and current employment are predictors of problem-focused coping, while psychological assistance is a predictor of emotion-focused coping. Avoidant coping showed no significant model fit. Conclusion: The findings underscore the critical need for comprehensive, evidence-based mental health interventions and accessible psychosocial support for Ukrainian refugees. The results indicate the modifiable post-migration factors important for refugee adaptation. Policy priorities should include early mental health screening targeting younger adults and unpartnered individuals, facilitation of skills-matched employment, provision of clear information about legal rights and available services, and culturally sensitive interventions that promote adaptive coping mechanisms.</p