18736 research outputs found
Sort by
Graph Neural Networks Powered Scientific Paper Recommendation
No full textScientific paper recommendation systems aim to help researchers discover relevant papers amidst the vast and ever-growing body of literature. With the exponential yearly increase in scientific publications, the demand for effective paper recommendation solutions has become both critical and increasingly challenging. In recent years, deep learning techniques have revolutionized recommender systems, and scientific paper recommendations have naturally integrated these advancements. In this dissertation, we address these challenges through three progressive contributions.
First, we enhance traditional content-based methods using Graph Neural Networks (GNNs) by introducing a Graph Convolutional Network-strengthened Topic Modeling (GCN-TM) approach. This method improves upon conventional topic modeling techniques by incorporating community insights through embedding propagation, enriching topic representations, and enabling more robust, community-aware recommendations.
Next, we address the dynamic nature of citation networks and their influence on paper embeddings as they evolve over time. To tackle the challenge of recommending relevant papers for works-in-progress, we develop a time-series embedding technique that captures shifts in citation network interactions. This method ensures the recommendation system remains adaptive and contextually relevant, even as the citation landscape changes.
Finally, we present a graph learning-based context-aware citation recommendation mechanism that models the decision-making process behind citation selection. To ensure scalability, we construct an overlay citation context network atop the traditional paper-to-paper citation graph. This overlay network acts as a guiding agent, facilitating accurate and context-sensitive citation recommendations during the composition of scientific documents.
This dissertation primarily focuses on scientific paper recommendation using graph learning techniques across multiple dimensions, including a hybrid approach to topic modeling for paper recommendation, temporal graph learning, and graph-based context-aware citation recommendation. The proposed methods not only enhance the precision and adaptability of recommendation systems but also provide practical guidance for both research and engineering applications. This work establishes a foundation for future innovations in scientific paper recommendation, equipping researchers with powerful tools to navigate and contribute to the ever-expanding body of scientific knowledge
Discrete Element Method Investigation Of The Dynamic And Seismic Response Of Embedded Geotechnical Systems
No full textThis study examines the seismic response of various embedded geotechnical systems utilizing the discrete element method (DEM). Embedded geotechnical systems play a crucial role in enhancing the stability and resilience of various structures as they provide solid foundations for infrastructural elements such as buildings, bridges, and transportation networks. The behavior of various embedded systems such as tunnels, pier foundations, and wind turbine foundations is investigated when subjected to seismic loading. In the first study, the seismic response of tunnels constructed in a deposit of granular soil of depth greater than 20 m is examined. Three-dimensional discrete element method simulations were used to model the soil assembly and the tunnel lining. Soil particles were simulated as rigid spherical particles that are allowed to overlap at the contact points
Intelligent Resource Allocation for SDN/NFV-Enabled Networks through Reinforcement Learning
No full textSoftware-Defined Networking (SDN) and Network Functions Virtualization (NFV) are two emerging paradigms that enable the feasible and scalable deployment of Virtual Network Functions (VNFs) in commercial-off-the-shelf (COTS) devices, which deliver a range of network services with reduced cost. The deployment of these services requires efficient resource allocation that fulfills the requirements in terms of Quality of Service (QoS) and Service-Level Agreement (SLA) while considering the constraints of the underlying infrastructure, such as maximum latency tolerance and affinity policies.
An optimized resource allocation result can benefit the network in various aspects, such as energy-saving, performance boost, and latency reduction. To achieve a fast, scalable, and dynamic composition of network functions to execute network services, we must first address the resource allocation problem in SDN/NFV-enabled networks, which involves numerous optimization variables resulting from the multidimensional space of system component parameters and states. It is especially nontrivial to address in a complex network as it involves a vast number of optimization variables resulting from the multidimensional space of network component parameters and states. Accordingly, determining the optimal resource allocation is an important and challenging problem to examine in SDN/NFV-enabled networks.
Reinforcement learning is a machine learning branch concerned with how intelligent agents ought to take actions in an environment to maximize the notion of cumulative reward. The agent can learn from the environment without the prior knowledge required and has the potential to outperform any expert. A thoroughly trained agent can make a decision that closely approximates real-time responsiveness. This capability is particularly beneficial in the context of SDN and NFV environments, where the ability to rapidly adapt to changing network conditions, optimize traffic flow, and efficiently allocate resources is critical for maintaining high levels of performance and reliability.
This dissertation contributes significantly to the field of network management by proposing versatile and robust solutions to resource allocation challenges in SDN/NFV environments. Our research paves the way for future advancements, offering a scalable framework that can adapt to the ever-changing dynamics of network technologies and demands. As networks continue to evolve, the principles and methodologies developed in this study hold the promise of shaping the future of network resource management, ensuring that SDN and NFV can fully realize their potential in creating more efficient, reliable, and high-performing network infrastructures
Statistical Approaches For The Early Detection Of Colorectal Cancer Using Longitudinal Biomarkers
Get PDFColorectal cancer (CRC) is the third leading cause of cancer-related death in the United States [45]. CRC is believed to advance from adenomatous polyps creating a unique opportunity for both early detection and cancer prevention [4, 23]. Like other diseases, CRC screening reduces mortality by detecting cancer at earlier, more treatable stages; however, it can also reduce incidence through the removal of precancerous lesions [4]. As a result, screening is recommended for average-risk adults ≥ 45 years of age and includes a variety of tests [4, 12]. Despite alternate screening options, colonoscopy capacity is often cited as a barrier to colorectal cancer (CRC) screening [28, 39, 44]. In this dissertation, we address capacity as a statistical problem rather than a resource one.
In the first part, we apply methods developed to incorporate longitudinal biomarkers for ovarian cancer screening to the data accumulated through a large FIT-based CRC screening program. This requires us to consider multiple methods to accommodate the necessary data transformation given the range for quantitative fecal hemoglobin concentration.
The second part of the dissertation looks at a new diagnostic marker obtained by extracting information from the biomarker trajectories using functional data analysis. The approach addresses problems of missing data and verification bias. Performance however is hindered by data sparsity which can be attributed to the screening process.
The third part of the dissertation revisits the method highlighted in part one to derive and evaluate a decision threshold for clinical implementation
Treaties Establishing ICAO And IMO – A Comparative Study
Get PDFThe comparison between air law and maritime law reveals both similarities and distinctions rooted in the unique frameworks of the International Civil Aviation Organization (ICAO) and The International Maritime Organization (IMO). While both entities were established through separate treaties, the Chicago Convention birthed ICAO, emphasizing the organization’s Assembly, Council, and auxiliary bodies. In contrast, the IMO Convention, also known as the Convention on the International Maritime Organization, forms the basis for IMO’s structure and functions as outlined in its preamble. The core objectives of IMO revolve around fostering collaboration among governments to enhance regulatory frameworks for international maritime trade. This encompasses advocating for elevated standards in maritime safety, navigation efficiency, and marine pollution prevention. Similarly, ICAO aims to establish principles and techniques for air navigation, promoting safe, regular, economical, and efficient air transport. The Chicago Convention primarily focuses on regulating international civil aviation, ensuring its orderly development and safety through defined principles and procedures, including standards for airspace sovereignty, aircraft registration, airworthiness, and aviation security. Conversely, the IMO Convention tackles various aspects of international maritime transportation, spanning safety, security, environmental protection, and shipping efficiency. Despite being specialized agencies of the United Nations, both ICAO and IMO face the challenge of accommodating diverse interests and viewpoints of their member states without the autonomy enjoyed by the private sector. Nonetheless, both organizations have consistently served the international community in facilitating world trade and commerce within their respective domains. This article discusses details of comparison and contrasts between ICAO, IMO, and air law and maritime law in their treaty settings
Multi-Class Emotion Classification with XGBoost Model Using Wearable EEG Headband Data
Get PDFElectroencephalography (EEG) or brainwave signals serve as a valuable source for discerning human activities, thoughts, and emotions. This study explores the efficacy of EXtreme Gradient Boosting (XGBoost) models in sentiment classification using EEG signals, specifically those captured by the MUSE EEG headband. The MUSE device, equipped with four EEG electrodes (TP9, AF7, AF8, TP10), offers a cost-effective alternative to traditional EEG setups, which often utilize over 60 channels in laboratory-grade settings. Leveraging a dataset from previous MUSE research (Bird, J. et al., 2019), emotional states (positive, neutral, and negative) were observed in a male and a female participant, each for 3 minutes per state while watching movie scenes designed to stimulate emotions. The dataset comprises 2548 features extracted statistically from each sliding time window (mean, median, standard deviation, etc.). Employing XGBoost, a subset of the top 100 features is selected from the original 2548, achieving an exceptional accuracy of 99.1%. This research aims to make significant contributions to accurately classify human emotion while advancing EEG-based sentiment classification for future real-time emotion prediction applications
Correcting Federal Rule of Evidence 404 to Clarify the Inadmissibility of Character Evidence
Get PDFCourts misinterpret Federal Rule of Evidence 404(b)(2) as an exception to Rule 404(b)(1)’s prohibition on character evidence rather than a mere clarification that emphasizes the permissibility of other-acts evidence whose relevance does not rely on propensity reasoning. This misinterpretation turns the rule against character evidence on its head by effectively replacing Rule 404 with a Rule 403 balancing—and one that incorrectly treats character inferences as probative rather than prejudicial, thereby favoring admissibility rather than exclusion. Consequently, as currently interpreted, Rule 404(b)(2) generates substantial unpredictability and verdicts based on conduct not at issue in a case.
I therefore propose that the Advisory Committee on Evidence Rules amend Rule 404(b)(2) to clarify the meaning of this rule as permitting only other-acts evidence whose relevance does not rely on a character inference— that is, whose chain of inferences is free of propensity reasoning. I show how the Advisory Committee can restore Rule 404’s logic and effectiveness through a straightforward modification in the language of Rule 404(b)(2). I then address the doctrine of chances—which pertains to a uniquely probative form of character evidence offered to prove the absence of chance or accident—and I explain why it should not cause reluctance to adopt my primary proposal. Then, as a secondary proposal (not required for the adoption of my primary proposal), I recommend amending Rule 404(b)(2) to establish a limited exception to Rule 404 for this type of evidence. I argue that my proposals to amend Rule 404(b)(2) would restore Rule 404’s meaning and intention to exclude evidence whose relevance relies on character reasoning and, in turn, would create fairer and more accurate trials
Testing the Effect of Autonomy-Supportive Instructions During Yoga on Autonomy, Competence, and Affective Response
Get PDFThis early-phase intervention development project tested the effect of autonomy-supportive instructional cues during a single yoga session on affective response, perceived autonomy, and perceived competence. Using a between-subjects experimental design, participants were randomly assigned to either an autonomy-supportive intervention condition or a mindfulness-based control condition. During the 30-minute online pre-recorded yoga sessions, affective response was measured immediately before the yoga session, at peak, pre-savasana, and post-savasana. Perceived autonomy and perceived competence were measured immediately after. Multilevel models tested the effects of the autonomy-supportive intervention on primary outcomes (affective response, perceived autonomy, and perceived competence) and secondary outcomes (yoga practice intentions and self-reported yoga behavior, considering the potential moderation by yoga experience. The intervention increased perceived autonomy but did not lead to more positive affective response or higher perceived competence. Unexpectedly, the control group reported higher intentions to practice yoga, but this did not translate into increased practice. Further research is needed to identify the types of yoga instructions that can decrease barriers and increase engagement in yoga practice
Maximizing Legged Accelerations: A Matter of Force, Time, and Gravity
Get PDFSprint running accelerations require runners to apply surface forces that: support body weight by pushing downward, accelerate the body horizontally by pushing backward, and align the direction of the push with the body’s mass center to maintain balance and posture, which imposes an upper limit on the average forward acceleration force equal to the average gravitational force (1.0 G) acting on the runner. This expectation arises from the mechanical constraints imposed by the need to generate sufficient vertical force to support body weight against gravity while simultaneously producing horizontal force to accelerate forward and aligning the push through the center of mass for balance. We tested the 1.0 G hypothesis by acquiring single-second sprint-start data from humans and canine sprinters in competition or equivalent (n=4 each). Additionally, we evaluated single-push data from human sprinters using force-instrumented (n=28) or platform-mounted (n=25) starting blocks against a condition-specific, single-push theorized limit of 1.25 G. The overall single-second race-start acceleration means of the human and canine sprint group (6.79±0.87 m•s-2, n=8) were significantly less than the theorized maximum of 9.81 m•s-2. Quadrupeds demonstrated a higher mean acceleration compared to bipeds (7.4±0.39 vs. 6.16±0.39 m•s-2). The single-push, mass-specific horizontal force maximums measured for human sprint athletes (0.99±0.06 G) also did not exceed the theorized gravitational limit. These results support our hypothesis that sprint acceleration maximums are imposed by gravitational forces and indicate that quadrupeds operate closer to this earthly limit than bipeds
The Blessed Assembly: Irreplaceable Physical Co-presence in Worship and Healthy Hybridity Reimagined after the Pandemic in the Digital Age
Get PDFBecause of the unprecedented and unexpected force of the pandemic since 2020, most churches around the world have experienced some online worship during the lockdown of their cities or the mandated closure of the church buildings. For many people, online worship seems to be an equivalent, if not better, alternative for gathering together—a physical co-presence in worship—even after the pandemic has ended. As necessary and vital as online worship experiences have been for Christians during the pandemic, the witness of the church from Pentecost throughout Christian history indicates that gathered worship in physical spaces is irreplaceable for faith formation and the embodiment of the Christian community. Seeing, hearing, reciting, singing, and moving with others while offering praise, confession, intercession, thanksgiving, dedication, and receiving God’s Word in a physical space are unifying worship acts indeed. Week after week, the worship actions are stamped and sealed in our memory, shaping us to live as God’s people together. This thesis explores and analyzes the potential and perils of online worship, our pastoral response to the hybrid life, the benefits of in-person embodied worship, and wisdom from the hybridity of the workplace to suggest a reimagined healthy hybridity for worship and the other ministries of the church. Nine aspects of corporate worship are advocated for the renewal of worship. Advice for pastoral care for online worshippers, guidelines for joining online worship, and a Trio Digital Detox practice are also highlighted at the end