116320 research outputs found
Sort by
Improving efflorescence resistance of metakaolin-based geopolymer via magnesium incorporation
No full textEfflorescence compromises the durability of geopolymers as it leads to excessive leaching and carbonation of alkalis. Unlike previous research that mitigates efflorescence of geopolymers mainly by reducing their permeability, this study aimed to stabilize the geopolymer framework by introducing magnesium ions (Mg2+). Upon the accelerated efflorescence, Mg2+-bearing geopolymers exhibited a higher retention of compressive strength. Such mitigation of efflorescence is unlikely to be associated with any changes in the crystallization pressure or pore structures, as samples with and without Mg2+ exhibit comparable pore characteristics. To the contrary, Mg2+, when dosed at 5 mol% and 10 mol% relative to the Na2O, substantially enhanced the stability of geopolymer gels, as characterized by a respective reduction of 25.3 % and 30.9 % in dealumination during leaching tests. Further analysis of the synthesized Mg-substituted sodium aluminosilicate hydrates (N-(M)-A-S-H) gels indicates that the polyvalent Mg2+, as compared to Na+ or K+, exhibits a higher binding energy when balancing the negative charges in geopolymers, thereby stabilizing their structures upon the efflorescence-induced attack. These findings offer a promising strategy to mitigate efflorescence from a nanostructural perspective
Message from the Program Co-Chairs: AIxVR 2025
No full textWelcome to the 7th IEEE International Conference on Artificial Intelligence & eXtended and Virtual Reality
RSVLM-QA: A Benchmark Dataset for Remote Sensing Vision Language Model-based Question Answering
Get PDFVisual Question Answering (VQA) in remote sensing (RS) is pivotal for interpreting Earth observation data. However, existing RS VQA datasets are constrained by limitations in annotation richness, question diversity, and the assessment of specific reasoning capabilities. This paper introduces Remote Sensing Vision Language Model Question Answering (RSVLM-QA) dataset, a new large-scale, content-rich VQA dataset for the RS domain. RSVLM-QA is constructed by integrating data from several prominent RS segmentation and detection datasets: WHU, LoveDA, INRIA, and iSAID. We employ an innovative dual-track annotation generation pipeline. Firstly, we leverage Large Language Models (LLMs), specifically GPT-4.1, with meticulously designed prompts to automatically generate a suite of detailed annotations including image captions, spatial relations, and semantic tags, alongside complex caption-based VQA pairs. Secondly, to address the challenging task of object counting in RS imagery, we have developed a specialized automated process that extracts object counts directly from the original segmentation data; GPT-4.1 then formulates natural language answers from these counts, which are paired with preset question templates to create counting QA pairs. RSVLM-QA comprises 13,820 images and 162,373 VQA pairs, featuring extensive annotations and diverse question types. We provide a detailed statistical analysis of the dataset and a comparison with existing RS VQA benchmarks, highlighting the superior depth and breadth of RSVLM-QA's annotations. Furthermore, we conduct benchmark experiments on Six mainstream Vision Language Models (VLMs), demonstrating that RSVLM-QA effectively evaluates and challenges the understanding and reasoning abilities of current VLMs in the RS domain. We believe RSVLM-QA will serve as a pivotal resource for the RS VQA and VLM research communities, poised to catalyze advancements in the field. The dataset, generation code, and benchmark models are publicly available at https://github.com/StarZi0213/RSVLM-QA
Hydrothermal liquefaction of sewage sludge: A comprehensive review of biocrude oil production, byproducts valorization, and future perspectives
Get PDFClimate change is driving global efforts toward carbon neutrality and expanding renewable energy sources. Hydrothermal liquefaction (HTL) of sewage sludge offers a promising pathway for sustainable biocrude oil production. This review systematically analyzes 956 records from Web of Science and Scopus databases, with 179 articles selected for detailed analysis following PRISMA guidelines. It presents the first comprehensive and systematic analysis of biocrude oil production and byproducts valorization from the HTL of sewage sludge. Key findings highlight that mixed sludge, with a balanced organic matter composition, is ideal for biocrude oil production, achieving an average yield of 38.95 % (range: 35.3–42.6 %). Higher biocrude oil yields are more likely to be achieved under reaction conditions of approximately 350 °C and a holding time of 30 min, as indicated by 2D kernel density estimation of the collected literature. These optimal conditions are summarized as a reference point for future studies, although the exact operating conditions may need specific exploration depending on the sludge properties. The transformation of organic matter follows the order: lipids > proteins > carbohydrates > lignin/humic substances, with diverse complex reactions driving biocrude oil formation. The biocrude oil contains significant heteroatom content-nitrogen (5.5 %, range: 0.23–9.3 %), sulfur (0.9 %, range: 0–4.3 %), and oxygen (15.7 %, range: 6.7–62.8 %)-which necessitate upgrading for biocrude oil applications. Nitrogen primarily distributes into the aqueous phase, while phosphorus and metals accumulate in the solid phases, offering opportunities for resource recovery. HTL also generates byproducts in aqueous (36.67 %, range: 0.19–60.3 %), solid (22.03 %, range: 0.43–50.73 %), and gaseous (13.71 %, range: 0.2–64.68 %) phases, which can be effectively valorized through proper management, promoting both industrial applications of HTL and the development of a circular economy. This work serves as a valuable resource for researchers, policymakers, and industry stakeholders, providing insights into biocrude oil production and byproduct utilization, advancing sustainable sludge management toward global carbon neutrality goals
Modelling Pavlovian biases in depressed and healthy young adults.
Get PDFPavlovian stimuli signalling potential punishment and reward have powerful effects on instrumental behaviours. For example, a cue associated with punishment will suppress well-learned instrumental responses. However, the degree to which Pavlovian stimuli interfere with the learning of instrumental responses is less well studied. In the current set of studies we investigated the effect of Pavlovian stimuli on instrumental learning and the extent to which depressive symptomatology moderated this relationship. We conducted two experiments using a sample of healthy adults and leveraged computational modelling to estimate learning parameters and the moderating role of depression on these learning parameters. In line with previous literature, participants found it more difficult to learn to make instrumental go and no-go responses in the presence of incongruent cues-for instance, making a "go" response for a cue which signalled punishment, and vice versa. Contrary to expectation we did not observe a reliable relationship between performance and depression scores; while Experiment 1 observed a relationship between depression and model-derived learning rates, these results were not replicated in Experiment 2. We discuss both the theoretical and practical implications of these findings in the General Discussion
Prospects of novel drug delivery systems in treating cerebral palsy
No full textCerebral palsy (CP) is a chronic disorder of children which have several etiologies. It is characterized by deficits in motor, sensory and cognition which occurs due to the injury to developing brain in uterus or soon after birth. Irrespective of several etiologies, neuro inflammation plays an important role in pathophysiology of brain in CP. About 2-3 cases per 1000 lives were found to be prevalent. The two variables that can be associated to CP are gestational age and birth weight. The prenatal, perinatal, preconception and postnatal categories are the identified risk factors. There is currently no effective treatment for CP, however drugs like anticonvulsants, antidepressants, antiinflammatories and benzodiazepines which offers their own characteristic side effects. On the other hand, phytoconstituents from ginger, astragalus etc., can also be used for treating CP which poses limitations like poor solubility, difficult to standardize etc. The development of remedies for the prevention and treatment of CP causing brain injuries opens up new possibilities with novel drug delivery systems (NDDS). Diverse systems of NDDS like nanoparticles, dendrimers, carbon nanotubes have been formulated to overcome the above said limitations of both synthetic and herbal drugs. This chapter focusses on various NDDS formulated against CP, their methods of preparations advantages and limitations
Embedding the SDGS in Teaching and Learning at Scale Across a University
No full textThe United Nations 17 Sustainable Development Goals (SDGs) set a framework for the key areas where action is needed to mitigate the impacts of climate change and address the key social, economic, and environmental issues facing the world today. In the Built Environment, educators have been teaching about environmental buildings and sustainability since the beginning of the 1990s. Since then, awareness of the issue has grown substantially, and sustainability education is delivered at undergraduate and post postgraduate across multiple disciplines. However, framing the education materials within the context of the 17 SDGs is less common or uniform. This chapter outlines how the 17 SDGs are being integrated in programmes across the University of Technology Sydney (UTS) in Australia, using examples from the undergraduate and post graduate Property and Planning courses
Facilitating Interdisciplinary Collaboration in Large-Scale Research Networks: Tackling Uncertainties in Knowledge Building and the Designing of Robotic Systems in Healthcare
No full textAbstract Robots used for care purposes have been the subject of considerable research effort, often interdisciplinary. However, our work has shown that there are frequently difficulties in working together to produce interdisciplinary knowledge on human robot and human computer interaction. We describe an initiative to enable participants within and across research projects to improve collaboration. Mandated by the funding body, we developed the Research Practice Workshop tool to bring together 10 projects and more than 100 people from science, technology and healthcare in a large-scale funding stream. Our tool creates space for interdisciplinary participation, despite its top-down project structure. We present data from seven Research Practice Workshops that illustrate how knowledge sharing worked and what challenges we faced in accordance with our iterative and participatory approach. The aim is to show how networks of practice transfer knowledge, facilitate interdisciplinary cooperation and create imaginaries of robotics for care. The Research Practice Workshops are a valuable tool for creating shared visions and practical implementations with all stakeholders in a democratic process. This demonstrates how design knowledge can become relevant in interdisciplinary collaboration and how Research Practice Workshops can support large scale research networks. Ultimately, our paper contributes both to best practices in heterodox scientific project management and to an understanding of how disciplinary perspectives mediate research
FedHigh: A Graph-Guided Aggregation Framework for Layer-Wised Personalized Federated Learning
Get PDFThe goal of Personalized Federated Learning (PFL) achieves model personalization on statistical heterogeneous scenarios. Existing PFL approaches roughly treat the entire model parameters as a whole unit for generating personalized weights, and neglect the latent correlations among parameters of different client models, which is crucial for model personalization by the same learning tasks on statistical heterogeneity. To solve this issue, we propose a graph-guided aggregation framework for layer-wised PFL, namely FedHigh. The framework leverages layered knowledge sharing of client models to generate latent correlation graphs in the hyperbolic space. The latent correlation graphs can efficiently guide the personalized aggregation. Experimental results demonstrate that FedHigh achieves significant accuracy improvements up to 42.6%, 33.1%, and 7.5% on graph, computer vision (CV), and natural language processing (NLP) datasets, respectively. Meanwhile, FedHigh accelerates convergence speed up to 52.6% compared to state-of-the-art baselines on CV datasets. Additionally, the scalability of FedHigh also attractively outperforms other baselines with different world sizes
Data sources:
Access Repository Dashboard
Do you manage OPUS? Access insider analytics, issue reports and manage access to outputs from your repository in the CORE Repository Dashboard!