Mason Journals (George Mason Univ.)
Not a member yet
3256 research outputs found
Sort by
An in-silico based approach to the analysis and determination of de novo peptide sequences and PEAKS efficacy based on LC-MS/MS data
No full textCurrently embargoed. Coming soon
Effective Machine Learning Algorithms Using Procrastination Indicators to Predict Task Completion Rates – A Literature Review
No full textAcademic procrastination is an issue many students face, and it can have consequences to students’ mental health and quality of work. Machine learning algorithms can be used to predict the effects of academic procrastination on task completion; however, these algorithms need features that are good indicators of procrastination. Procrastination indicators used by previous studies include student self-ratings of academic procrastination, activity logs, time-stamped trace data of studying, and interaction behavior of computer-based learning environments. Previous studies explore the use of new technology, apps, and virtual learning assistants to monitor and predict procrastination-related behaviors. These studies use techniques such process mining, unsupervised machine learning, and probabilistic mixture models to identify student behavior patterns and connection to academic outcomes. For identifying the relationship between procrastination and academic performance, qualitative data (e.g., detailed student activity logs) was found more valuable than quantitative data (e.g., timestamped interactions, with computer-based learning environments). The next step of this research is to design an experiment and collect student procrastination indicator data that can be used to develop a random forest algorithm to predict task completion rates
Comparing Transformers & Attention Mechanisms, Generative Adversarial Networks, and Reinforcement Learning to Human Learning Processes – A Literature Review
No full textRecent advancements in Artificial Intelligence (AI) have introduced powerful machine learning techniques, such as Transformers & Attention Mechanisms, Generative Adversarial Networks (GANs), and Reinforcement Learning (RL). Comparing these methods with human learning processes provides insights into their strengths and limitations. Transformers & Attention Mechanisms demonstrate human-like selective attention, effectively comprehending the broader context while focusing on essential details, reflecting cognitive processes in humans. However, the inherent differences between AI and human attention mechanisms highlight distinct approaches: algorithmic processing in AI versus human cognition. GANs are capable of generating synthetic data, showcasing creativity in producing realistic artifacts, but their algorithmic nature limits their depth compared to human artistic creativity. To address this limitation, researchers have explored introducing "arousal potential" to encourage deviations from established styles in art generation, enhancing the artistic appeal of GANs' output. RL, mimicking human learning in sequential decision-making tasks, often outperforms human strategies, indicating its effectiveness in learning complex strategies and exploring solutions more effectively than humans. Although AI still lacks the intrinsic depth, intuition, and creativity found in human cognition. Understanding these similarities and distinctions can guide the future development of more sophisticated and human-like AI systems, facilitating progress in AI research and applications. and applications.
 
Analysis of Spatial-Temporal Trends in Vegetation and Rainfall in the Sahel Region during 2000-2020
No full textClimate change has a profound impact on the natural resources in the Sahel region. To understand vegetation dynamics and precipitation changes, this study examines the spatial-temporal trends in the Sahel over the years 2000-2020. The MODIS Normalized Difference Vegetation Index (NDVI) and Merged Satellite-Gauge Precipitation Estimate (IMERG) datasets are utilized, employing linear regression and Pearson correlation analysis to assess the relationship between vegetation index and rainfall, and explore seasonal correlations to understand variations during the growing season (July-October). The results show an increase in average annual NDVI (0.0005 yr-1, p < 0.05) and precipitation (0.0003 mm/hr/yr, p < 0.05). NDVI (0.0009 yr-1, p < 0.05) and precipitation (0.0009 mm/hr/yr, p < 0.05) in the growing season also show a rise. The south-central Sahel experiences the largest NDVI increase (11.66%), while the northeast sees the highest precipitation rise (125.15%) from 2000 to 2020. Additionally, a correlation map between rainfall and NDVI highlights that during the growing season, 18.505% of pixels exhibit a positive correlation above 0.7 with the southwestern part showing the highest correlation. These seasonal correlations shed light on the intricate interactions between NDVI and rainfall in different periods, offering valuable insights for ecosystem management in the Sahel. This study provides valuable support for conservation efforts, highlighting the importance of understanding these trends urgently to safeguard the region’s natural resources and bolster the Africa-Water-Energy-Food-Health (AWEFH) Nexus
Do YouTubers Promote Bullshiting using ChatGPT? Exploring the Use of Large-Language Models in YouTube Videos and Their Risk Landscapes
No full textLarge language models (LLMs), such as ChatGPT, are by nature “bullshit” generators, as they are not recognizant of “truths.” As per Frankfurt’s theory, bullshit is defined as utterances made without acknowledgment of the truth. LLMs generate responses by predicting the next most probable word in a sequence without recognition of the truth. As ChatGPT peaks in the hype cycle as of early 2023, there have been many media contents that cover ChatGPT and similar LLMs. At best, these media contents can be beneficial in helping people understand the technology; but at worse, they could facilitate the generation of bullshit on the Internet, which is already filled with human-generated misinformation. This makes it necessary to understand how much the media contents have potential risk (or provide opportunities) to the media environment on the Internet. To understand the use of ChatGTP and their potential risk of bullshit generation, we analyzed YouTube videos that have been published since the introduction of ChatGTP in December 2023. Among the entire videos collected from YouTube APIs using keywords, “ChatGPT, GPT-3.5, and GPT-4,” we randomly sampled about 400 Youtube videos and analyzed their potential risk of bullshit, manually categorizing videos as low risk, high risk, reducing risk, or bullshit itself. A majority of high-risk videos were about how to make money through automatically generating media content, whereas experimental videos that tested ChatGPT’s capabilities tended to reduce the risk of bullshit by explaining the shortfalls of ChatGPT. In addition, sentiment analysis of the videos’ comments, ANOVA, and Fisher’s exact test were conducted to examine the relationships between video features, type of bullshit risk, and video performance. Results show that high-risk YouTube videos may be incentivized more in spreading bullshit, as they promote monetary success using ChatGPT. Also, based on the baseline understanding of YouTube videos, we developed a machine learning model that can predict videos’ bullshit risks by leveraging channel statistics such as the number of views, likes, comments, and subscribers. Our study provides an implication for social media designers that, as there is a high potential of AI-generated content to spread bullshit, there is a need to develop more effective content moderation strategies
Determining and Applying the Relationship Between FWI and FRP In Predicting Fire Change
No full textFire Weather Index (FWI) analyzes how fire-prone an area is through considering the Initial Spread Index, or the wind and moisture levels in the area, and Buildup Index, or the amount of fuel available for combustion. An understanding of the relationship between FWI and changes in Fire Radiative Power (FRP), or the amount of energy released by fires, has been sought after in order to better fire management/prevention efforts. While FWI can analyze how susceptible an area is to fires, we lack an understanding of how FWI can be applied to predict fire change. Using 2020’s fire data, we analyzed fire duration and FWI distribution through scatterplots and histograms. Using this analysis, we performed statistical normalization to create a CDF of normalized FWI values on a 0 to 1 scale, removing the bottom and top 10% of FWI data in order to truncate out any extreme outliers. Our method involved normalizing a day’s FWI value, finding its corresponding value in the CDF, and multiplying this newfound value by a day’s FRP to find the next day’s FRP. This method was used to perform both 7-day and next-day forecasts, comparing predicted FRP to actual FRP. I focused on the specific region of the August Complex Fire, where my predictions proved to have a lower error than assuming a 25% decrease in FRP or persistent FRP, with the error only decreasing as the forecasts grew longer. The implications therein remain that FWI predicts FRP change more accurately than assuming no change or assuming a constant change
Binding of indolicidin to the DMPC bilayer using molecular dynamics simulations
No full textAntimicrobial peptides (AMPs) are part of the innate immune system and fight bacterial infections by disrupting bacterial lipid bilayers or encouraging water permeation. Indolicidin, a cationic AMP with 13 amino acids, has garnered attention due to its broad-spectrum activity, unique secondary structure profile compared to other AMPs, and short sequence. We used molecular dynamics simulations to study the binding of indolicidin peptides to the DMPC bilayer. We analyzed this system by extracting the atomic coordinates as a function of time and computing the distance between indolicidin and the DMPC bilayer. We also performed secondary structure analysis to understand the effect of binding on indolicidin structure. Overall, our molecular dynamics simulations assess the physicochemical mechanisms of indolicidin’s initial docking process to the DMPC bilayer. In future studies, we will extend these simulations to establish the impact of indolicidin’s equilibrium binding profile on its antimicrobial activity. 
Application of Nylon Affinity Nets to Capture Tuberculosis Extracellular Vesicles and Identify Protein Biomarkers
No full textTuberculosis (TB) requires efforts for more efficient diagnostic approaches to combat its impact worldwide. Current diagnostic efforts only identify active disease rather than infection stage and are limited to blood or skin samples, which are difficult to obtain from younger or immunocompromised patients. A promising source for TB diagnosis is the extracellular vesicles (EVs) released by TB bacteria, containing valuable TB biomarkers in many easily accessible biological samples. They provide an accurate distinction between latent TB and TB disease, and thus, can be more informative if captured properly. This study develops a novel method to utilize nylon affinity nets to properly capture EVs from patient urine samples. We used nylon fibers functionalized with synthetic dyes (affinity net) for the capture and concentration of EVs from urine. EVs are examined for protein and DNA content by LC-MS/MS and by identifying the TB genes within EVs using molecular biology techniques (DNA precipitation - ligation mediated amplification, cloning sequencing, and nucleotide blast). Functionalized nylon captured urine EVs contain MTB peptides and nucleic acids, such as LAM, CD81, CD63, CD9, and the rpoB gene (identified by PCR). Identification of various proteins and markers in TB urine could help in sputum-free diagnosis and antibody development. 
Deciphering Susceptibility Factors of MERS-CoV in Ethiopia: Insights from a Comprehensive Zoonotic Disease Survey
No full textMiddle East Respiratory Syndrome Coronavirus (MERS-CoV) is a zoonotic virus that has raised serious concerns as a public health threat worldwide. Ethiopia, with its substantial camel population and diverse wildlife, represents an important ecological context for studying the transmission dynamics of MERS-CoV between other animals and humans. However, the factors that trigger zoonotic disease emergence and spread remain poorly understood. Through analyzing and condensing data from a zoonotic disease survey, considering various factors such as occupation, drinking water sources, current health status, or number of animals owned, we are now able to identify which groups are most susceptible to these diseases due to different environments and styles of living. From the simple data analysis, we can reasonably infer that there are large numbers of older people, individuals with weakened immune systems, and poor health statuses. These groups are likely at greater risk of developing severe disease. Additionally, those who work around animals, eat animal meat, or use them for skins and medicine are more vulnerable to zoonotic diseases as well
Understanding the spread of racism through mathematical modeling, analysis and simulation
No full textRacial discrimination has continued to play a factor in the many lives of minorities in America. This research project aims to understand the passage of racism through different stages and illustrate how certain factors can minimize the spread of these ideals through mathematical modeling. The SIR (Susceptible-Infectious-Recovered) model, traditionally used in epidemiology to analyze the spread of infectious disease, is adapted to highlight the specific groups who interact with the spread of racism. The utilization of the SIR model in the context of racism allows for an alternate perspective to transmission of discriminatory beliefs and initiatives within social settings. The new model created was numerically solved and validated for various transmission parameters. The research will also address the role of bystanders in situations of racial bias, where individuals witnessing discriminatory acts possess the power to intervene, however, choose not to actively share their beliefs. The implications of this interdisciplinary research not only offers insightful information about the dynamics of discrimination using mathematical modeling but establishes the importance of responsibility to encourage action among bystanders further transforming the impact of societal attitudes and behaviors.