22505 research outputs found
Sort by
Desiccation Tolerance in Tetradesmus Green Algae
No full textIn the face of the global climate change threat, understanding the adaptations that organisms have evolved to handle environmental variation is of particular interest to scientists. With climate change impacting global water availability and increasing the risk of drought for many traditional agricultural areas in the United States (IPCC, 2014, IPCC, 2021, and Kuwayama et al., 2019), desiccation tolerance in vegetative states is one adaptation that is currently receiving a lot of attention. Green algae are particularly useful organisms for understanding this adaptation due to their ubiquity across environments, which has given rise to independently evolved organisms displaying different levels of desiccation tolerance. In comparing the differential gene expression in gene families traditionally linked to desiccation tolerance in green plants, such as Early Light Induced Proteins (ELIPs), Late Embryogenesis Abundant Proteins (LEAs), and Heat Shock Proteins (HSPs), across two closely related strains of algae-Tetradesmus deserticola and Tetradesmus obliquus, which display contrasting levels of tolerance to water loss- we aim to improve our understanding of the drivers of desiccation tolerance in plants
Simulation-Based Learning in Security Studies: A Rubric-Guided Approach to NATO–Russia Dynamics on the Korean Peninsula
No full textThis article examines the implementation and efficacy of a team-taught simulation-based learning series designed to deepen students’ understanding of NATO–Russia interactions on the Korean Peninsula. Drawing on the principles of Bloom’s Taxonomy and guided by a comprehensive rubric, this framework provides a structured, progressive approach to developing learners’ analytical, evaluative, and creative competencies. Through six sequential simulations using the SEERIST platform—culminating in strategic policy recommendations—students enhance their ability to recall foundational knowledge, apply complex concepts, analyze multifaceted scenarios, evaluate diverse outcomes, and propose innovative solutions. The article discusses theoretical underpinnings, the design and implementation of the simulation series, the associated rubric, and reflections on student outcomes and pedagogical implications
Using Generative Models for Data Augmentation in Medical Imaging
No full textObtaining large amounts of labelled data in medical imaging is hindered by issues like privacy and cost, therefore limiting the gains that can be achieved from Deep Neural Networks (DNNs). A lot of work has shown data augmentation to be an effective strategy for expanding training data leading to more robust models. However, the lack of data is not always uniform across datasets e.g., some classes in a dataset may be underrepresented. This is particularly problematic because the overall good performance of a model on a dataset can easily mask the under-performance of underrepresented classes or instances. Naively employing data augmentation in such scenarios can lead to more performance disparities. In this research, we focus on two such cases in which the overall dataset may not exhibit problems of data scarcity but some classes or instances are underrepresented. 1) When some objects are underrepresented in image segmentation e.g., tumours in medical images. 2) When a particular event is underrepresented for a population subgroup hence leading to spurious correlation in data. For the first case, we propose a novel object-centric data augmentation model that can learn the shape variations for the objects of interest and augment the object in place without modifying the rest of the image. For the second case (spurious correlations), in this proposal, we take the first step in systematically understanding the shortcomings of existing optimization and representation based approaches to tackle spurious correlations in the context of medical images. Our findings show that current optimization methods for dealing with spurious correlations based on underperforming samples can be problematic when bias is not the only cause for poor performance, and naive invariant representation learning suffers from spurious correlations itself. We further show that using optimization in conjunction with invariant representation learning can lead to better representations that are void of irrelevant features to the task at hand. We also try to tackle the problem of spurious correlations in the data space by exploring the possibility of using counterfactual (CF) augmentation to factor out correlations from the training set. However, despite being widely utilized in the literature, most CF generation models struggle when exposed to highly correlated data, and oftentimes strategies to improve the accuracy of CFs come at the cost of their diversity. Finally, in future work, we propose to look beyond the loss-based methods for the identification of correlation-conflicting samples, as the loss-based methods often cannot distinguish between noisy and correlation-conflicting samples
Solar Power Forecasting Using Deep Learning Techniques
No full textThe growing utilization of solar energy and other renewable energy sources has heightened the demand for a proper forecasting model that enables efficient control of the energy supply as well as grid reliability. In this work, we concentrate about the use of novel techniques in ML models, namely Deep Neural Networks (DNN) and Convolutional Neural Networks (CNN), to try to forecast power output in a solar-powered plant using past weather and power output records. Production of solar energy is naturally volatile because it is dependent on the weather conditions and is problematic in nature when integrated into power grids. Conventional statistical forecasting models tend to miss the nonlinear and complicated manifestations between the environmental factors and the energy production. The study is rooted in the necessity to address these shortcomings by creating machine learning models capable of more dependably mirroring the trends in solar power production. The weather parameters that were used in the dataset are the ambient temperature, module temperature, and irradiation as three weather variables, whereas the parameter used in the data that were collected in a solar power plant was the DC power, AC power, daily yield, and total yield. There was a lot of preprocessing of data in the study to handle missing values to make the data consistent. These were DNN models to ensure that the complex non-linear dependency was captured, and CNN models that were used to recognize the spatial elements of the structured data. Principal research questions were aimed at discovering the key weather parameters that influence the production of solar power, as well as the determination of the suitability of DNN and CNN models to achieve correct forecasts. The data show that the accuracy of predictions using both models stands higher than traditional approaches, and CNN is better than DNN in terms of the error values (such as RMSE and MAPE). The study finds that deep learning s and specifically and ML models generally, is a feasible approach to improve solar power forecasting. Other areas for future research include hybrid models, real-time data integration and deployment in various locations all over the world to enhance the accuracy and usability of power components systems. Keywords: Solar Power Forecasting, Deep Learning, DNN, CNN, Renewable Energy, Energy Management, Machine Learning, Weather Data, Grid Reliability, Sustainable Energy
Machine Learning Approaches for Predicting and Preventing Vehicle Theft
No full textThis study develops a data-driven framework for predicting vehicle theft using machine-learning techniques applied to a comprehensive crime dataset containing demographic, spatial ,and operational policing variables. The dataset, sourced from publicly available crime records, included 9,712 cases and 30 variables, with minimal missing data and balanced class distribution. After performing extensive preprocessing—including outlier analysis, missing-value imputation, categorical encoding, feature selection, and correlation testing—four supervised classification models were implemented: Logistic Regression, Random Forest, Support Vector Machine, and a Neural Network. Bivariate statistical tests, including Chi-square and Mann–Whitney U, revealed significant relationships between theft occurrence and factors such as crime area, luxury vehicle density, police response time, and distance to the nearest police station. These insights guided the feature selection process before model development. Model performance was evaluated using accuracy, precision, recall, F1-score, ROC curve, and confusion matrices. Among the four models, the Neural Network demonstrated the highest predictive performance, achieving an ac curacy of 92%,alongside strong precision and recall values. The model showed a well-balanced ability to correctly classify both theft and non-theft cases while minimizing false negatives, which is operationally critical in crime-prevention contexts. ROC analysis further confirmed its robustness, with an AUC score near optimal discrimination levels. The findings highlight the effectiveness of machine-learning approaches—particularly Neural Networks—in forecasting vehicle theft patterns. The study provides a practical, interpretable, and operationally relevant framework to support law-enforcement agencies in resource allocation, hotspot identification, and strategic crime-prevention planning
PREDICTIVE AI MODELS FOR PHISHING ATTACK DETECTION: A DATA-DRIVEN AND STATISTICAL ANALYSIS APPROACH
No full textPhishing has remained a serious threat to cybersecurity, as this type of attack can easily bypass detection systems that are either rule-based or blacklist-based. The proposed thesis work will present a solution that will make use of both statistical analysis and machine learning to correctly identify a phishing website. The solution will make use of a hybrid approach comprising statistical-based preprocessing methodologies, such as PCA and decision tree-based feature selection, to filter the crucial URL features that a website may possess. A carefully balanced dataset has been utilized, as well as a non-parametric approach utilizing the Mann-Whitney U-test to validate if the features are statistically important. The thesis will conclude by building a neural network model capable of classifying a phishing website with accuracy through optimized URL features
Enhancing the Immunogenicity of Extracellular Vesicle-based Vaccines Against Pathogenic NTHi Bacteria
No full textAcute otitis media (AOM), or ear infections, are highly prevalent in children, with nearly 80% experiencing at least one AOM by the age of three. As such, there is an urgent need for a vaccine against Nontypeable Haemophilus influenzae (NTHi), a main cause of AOM in the US. Most vaccines contain multiple antigens from the pathogen, inducing an immune response and antibodies to those antigens. Here, we describe a novel approach to creating a vaccine against NTHi, incorporating a unique recombinant trivalent protein (Tri) featuring three NTHi antigens (OMP26, P6, and protein F) in a single peptide chain with an N-terminal signal sequence, contained within an extracellular vesicle (EV) produced by Escherichia coli (E. coli). The N-terminal signal sequence was added to this trivalent peptide to lipidate the trivalent protein and shuttle it to the E.coli membrane. We produced and characterized the Tri-EV to confirm the presence of all three NTHi antigens in the EV. Preliminary findings suggest that Tri-EV contains all three antigenic components, and vaccination with the Tri-EV produces antibodies to those antigens in mice. Continued work on this promising vaccine candidate will provide insight as we look to develop effective EV-based vaccines against relevant pathogens
Transgender Adolescents\u27 Healthcare Rights: The Gap Between Gender-Affirming-Care (GAC) Policymakers’ and Healthcare Providers’ Goals
No full textTransgender people have marginalized healthcare experiences that result in decreased health outcomes. For transgender adolescents, gender-affirming-care (GAC) is even more difficult to access and being restricted with increasing frequency. This mixed methods study was conducted to examine transgender healthcare through the gaps between GAC policymakers’ and healthcare providers’ goals. General research has focused on the relationship between GAC and transgender health, GAC policy development, and the discrepancies between GAC policy outcomes and medical professionals’ advice. This leads to the research question about how policymakers and healthcare providers can collaborate to bridge the gap between GAC policies and transgender minors\u27 healthcare experiences, ultimately improving the adolescents’ overall well-being. To understand the impact GAC policies have on transgender youth, data was compiled herein and analyzed to investigate the gap and the consequences of the gap between policymakers’ and healthcare providers’ goals. Quantitative data collection focused on health statistics by state, political stance and supported policies, and healthcare access. The case studies mainly focus on surveys and interviews. Qualitative data collection focused on transgender healthcare affected individuals’ interviews, emphasizing experiences with GAC policies and the healthcare effects. Findings reveal a positive correlation between transgender youths’ wellbeing and increased access to GAC. Additional findings reveal that anti-GAC states are more biased against transgender people. The statistical trends seen across multiple case studies dismantle anti-GAC policy advocates’ argument, highlighting the need to listen to medical professionals about transgender youths’ healthcare to improve the adolescents’ wellbeing
Hugs in Hues
No full textHugs in Hues is an animated short film about social anxiety, emotional expression, and discomfort with physical touch. The story follows Tina, a young artist who struggles to express how she feels. When she pulls away from a hug from her friend Emily, Tina escapes into a painting-like 2D world made from her drawings. This imagined world reflects Tina’s emotions. The colors, textures, and spaces change based on how she feels. Her anxiety, fear of rejection, and need for safety take visual form. The real world feels sharp and rigid, while Tina’s illustrated world feels soft, imperfect, and expressive. Animation allows these emotions to be shown visually, rather than explained through dialogue. As Tina moves through this inner space, she slowly faces her fears. She begins to understand that being vulnerable does not always lead to rejection. Through a mix of 3D and 2D animation, the film shows how inner emotional experiences can be made visible. Hugs in Hues highlights art as a way to cope, heal, and reconnect with others