62588 research outputs found
Sort by
Domain-Specific Dialogue: Fine-Tuning Large Language Models for Enhanced Research Outcomes and Insights - Project Summary
No full textFine-Tuning Large Language Models for Enhanced Research Outcomes and Insights with Domain-Specific DialogueLarge Language Models (LLMs) have become pivotal technology, enabling machines to understand and generate human-like replies to questions. Popular pre-trained LLMs have a general understanding of language and capture a wide range of linguistic patterns, but may not perform in specific tasks or domains. Specialized training or fine-tuning is needed to help improve their performance and accuracy. Fine-tuning LLMs allows us to customize the model to a specific domain, enable LLMs to better understand domain-specific terminology, jargon, and context, and could provide enhanced and expedited research outcomes and insights. This use-case will be designed to serve as a foundation for building a versatile documents-to-LLM pipeline, capable of being adapted and reused across various academic disciplines. In addition, education opportunities will be provided to participating students with in-depth knowledge of AI and LLMs, enabling them to develop expertise in designing, and applying these powerful tools to drive innovation and discovery across various academic disciplines
THREE ESSAYS IN HEALTH ECONOMICS
No full textThis dissertation comprises three essays, each exploring key issues in health economics and constituting a separate chapter. These essays collectively address critical dimensions of health inequality and patient behavior in the health market.Chapter 1: How, and how well, do patients choose quality providers? Evidence from heart surgery in Pennsylvania.Patient choices of health providers are built upon limited information due to information asymmetry. This chapter examines how patients select surgeons for CABG/heart valve surgeries when public information on surgeons\u27 performance is available. I use data from the Pennsylvania Health Care Cost Containment Council\u27s (PHC4) Cardiac Surgery report to estimate patients\u27 responses to reported surgeon performance using a discrete choice model, with an emphasis on differences by patient sex and insurance type. Patients exhibit a modest but noticeable response to quality information about surgeons, at least in the Central and Western regions of Pennsylvania. Female patients and Medicaid patients are found to be less influenced by this quality information and consequently are more likely to be treated by low-performing surgeons.Chapter 2: Does healthcare quality moderate the relationship between mortality and income in the United States?This chapter investigates whether healthcare quality moderates the strong relationship between mortality and income in the United States. I use county-level data on age-adjusted mortality and median household income on the white non-Hispanic group, along with a composite measure of hospital quality from CMS reports, to estimate regressions of mortality on income alone and then with hospital quality included. The quality of hospitals in a county does have an inverse relationship with mortality, but including this has almost no effect on the strong relationship between mortality and income. This chapter also examines the roles of education and health insurance coverage and finds that when education is included the association of mortality with income (at the aggregate level) is substantially reduced.Chapter 3: Healthcare expenditures among fathers with young childrenUnderstanding how fathers\u27 health expenditures are associated with the presence of a young child in the family is crucial for comprehending the impact of this significant life event. This chapter compares personal healthcare expenditures among fathers with young children and those with older children, utilizing data from the Medical Expenditure Panel Study (MEPS) from 2014 to 2018 and employing a two-part model. It reveals a strong relationship between personal healthcare expenditures and family income, particularly among fathers living with children under 2 compared to those with older children. On average, living with a child under 2 in the 25th income percentile family is correlated with lower total health expenditures by approximately for fathers, whereas no significant correlation was observed in families at the 75th percentile.</p
Temple University and its impact on the redevelopment of North Philadelphia
No full text{"value":"This dissertation examines Temple University as a case study to demonstrate the larger impact universities have had on their surrounding neighborhoods. Temple University’s founding mission vowed to provide an affordable education and help those in its “backyard” as a beacon of public service, social activism, and community engagement. Despite Temple’s founding mission and its status as a public university receiving public funds, it largely failed to meet its responsibility to its neighborhood. Temple, located in North Philadelphia, a predominantly low-income and Black neighborhood, experienced massive changes to the cultural, social and economic fabric of its community and thousands were displaced from their homes due to Temple’s expansion. This study is organized chronologically tracing Temple’s origins to its beginning expansion in the 1950s ending with the football stadium controversy in 2020.
","attr0":"abstract"
Man and/or Machine: Anxieties of Disability and Super-Ability in Nineteenth-Century Literature
No full text{"value":"There is no shortage of research on industrialization anxiety’s impact on nineteenth-century literature. What has been less thoroughly examined, however, is the
link between the anxieties of industrialization and disability, and how this link was
explored in Victorian-era texts. Yet, a disability studies approach to such literature proves
generative. The Industrial Revolution saw workers not only displaced, but also disabled
by machinery, and these two outcomes became intertwined in the Victorian imagination.
Contemporary texts, ranging from Edgar Allan Poe’s “The Man That Was Used Up” to
William Dodd’s Narrative, reveal that Victorians were deeply concerned by these threats
to their way of life. However, industrial technology also offered the intriguing potential
for what might be termed “super-ability.” Increasingly sophisticated prosthetics, medical
interventions, and scientific knowledge offered a taste of unprecedented power and
longevity. This, too, proved foreboding in its own way: If humanity was not replaced by
machines, it might instead merge with them. Such a concept once again posed a mortal
threat to humanity as the Victorians knew it; perhaps, critic Laura L. Behling posits, they
would discover that “the perfect human may not be human at all,” but rather a machine.
These conflicting concerns rendered the Victorians all the more anxious – and their
literature all the more fascinating. This paper examines the interplay among
industrialization, disability, and super-ability in British and American texts of the
Victorian era. In so doing, it seeks to provide a disability-informed perspective on the
impact of industrialization on Victorian-era literature, and in turn, the disabled
communities it depicted.","attr0":"abstract"
Defending Democracy: A Conversation with Liz Cheney
No full textLiz Cheney served as the U.S. representative for Wyoming\u27s at-large congressional district from 2017 to 2023. She chaired the House Republican Conference, the third-highest position in the House Republican leadership, from 2019 to 2021, and served as the Vice Chair of the Select Committee to Investigate the January 6th Attack on the United States Capitol. She was also a member of the House Armed Services Committee, China Task Force, Natural Resources Committee, and the House Committee on Rules.
Cheney served previously at the State Department as the Principal Deputy Assistant Secretary of State for Near Eastern Affairs, and in positions for USAID and the Department of State working in Poland, Hungary, Russia, and Ukraine. Cheney practiced law with White and Case and at the International Finance Corporation. A specialist in national security and foreign policy, she is the co-author, along with her father, former Vice President Dick Cheney, of Exceptional: Why the World Needs a Powerful America.
Cheney received her Bachelor of Arts degree from Colorado College, and received her Juris Doctor from the University of Chicago Law School. In 2022, Cheney, along with Ukrainian President Volodymyr Zelensky, received the the John F. Kennedy Presidential Library\u27s prestigious Profile in Courage Award, with a commendation for her "consistent and courageous voice in defense of democracy.
Physics-Guided AI Towards Better Diagnosis on Heart Diseases - Project Summary
No full textWe are developing physics-guided machine learning techniques for cardiovascular systems, to obtain a personalized digital twin model for heart diseases.Heart disease, characterized by changes in vascular, valvular, and ventricular systems, has been the leading cause of death in US since 1950. These diseases often happen acutely, making early detection essential. This interdisciplinary project aims to transform the clinical approach to early stage heart diseases, which is still dominant by traditional methods, through an AI-based approach. Our goal is to integrate data from wearable devices, mechanical measurements, and physics-based modeling, towards a personalized digital twin model for heart diseases. To this end, the project has three sub-tasks. First, a physically interpretable constitutive modeling approach will be developed, to capture the change of mechanical responses and the underlying mechanisms for cardiovascular tissue degradation. Second, an attention-mechanism-based approach will be designed, to infer the correlation between multiphysics data and the corresponding tissue model. Third, an efficient surrogate model will be constructed, enabling fast inference to assist in real-time monitoring. As the long-term goal, this personalized digital twin model is anticipated to provide early risk assessments, while also offering insights into cardiovascular mechanics that could lead to new treatment strategies.
In the past semester, our team has made progresses on the first and the second sub-tasks. For the first sub-task, a peridynamic neural operator model has been developed, which constructs the constitutive law as well as the underlying fiber orientation from loading-response measurements. To
demonstrate the applicability of our approach, we apply the HeteroPNO in learning a tissue model and fiber orientation field from DIC measurements of a heart valve leaflet specimen. For the second sub-task, we have developed an attention-mechanism-base approach to extract global prior information from mechanical measurements of multiple tissue specimens, and provide a generalizable foundation model to new and unseen tissues. As the ongoing work, we are extending this model to multiphysics and multimodal data
Using herbarium moss specimens to investigate testate amoeba ecology and responses to environmental changes in the New Jersey Pine Barrens.
No full textHerbaria, collections of pressed and dried plants, are a rich source of information for taxonomic, genetic, and ecological studies. They may also contain information on past environmental conditions and even the microbial communities that live on plants. For example, Sphagnum moss is well represented in herbaria, is an excellent collector of airborne pollution, and serves as a substrate for diverse microbial communities. I conducted a study to 1) assess the preservation and potential use of testate amoebae recovered from herbarium Sphagnum collections, 2) investigate testate amoeba ecology in the New Jersey Pine Barrens, 3) assess changes in testate amoeba community composition that may have occurred in response to heavy metal deposition, and 4) provide a record of heavy metal deposition using herbarium Sphagnum from the Pine Barrens. Testate amoebae were extracted and quantified from 161 herbarium specimens and 53 fresh Sphagnum samples, and community composition was compared with water-table depth measurements, the moisture ecology of Sphagnum species, the concentrations of ragweed pollen, Mn, Fe, Co, Ni, Cu, Zn, Cd, and Pb. Results indicated good preservation of testate amoeba community composition in herbarium mosses, including the preservation of siliceous tests that generally do not preserve well in bogs. Results also confirmed the primary importance of substrate moisture in controlling the composition of testate amoeba communities, and also suggested that landscape disturbance, as indicated by the concentration of ragweed pollen, affects the structure of testate amoeba communities. In drier habitats like hummocks, the composition of testate amoeba communities was correlated with the concentration of zinc, with higher zinc concentrations associated with a higher abundance of mixotrophic testate amoebae. The record of heavy metal deposition on mosses since the early-to-mid 1800s provides a detailed history of pollution that can be linked to industrial and other human activities, as well as the shift toward a preservation ethic since the 1970s. Higher median values of Ni and Cu occurred during the mid 1800s, when mining and iron smelting were widespread, while many heavy metals have decreased in concentration since the 1970s. This is particularly prominent in the record of Pb, which decreased after the phase-out of leaded gasoline. This work brings forward a novel use of herbarium specimens and highlights their potential contribution to understanding pollution history, microbial ecology and biogeography, and plant-microbe interactions, potentially leading to new ways of examining how and why microbial communities have changed over time and the effects of human activities.</p
Exploring the Synthesis – Structure – Property Relationship of Inorganic Nitride and Oxide Thin Films
No full textThin films of transition metal nitrides and oxides have a wide range of applications in electronic and optoelectronic devices, corrosion and wear resistant coatings, energy storage, catalysis, biomedical applications and so on. To develop new and improved materials by design, for such applications require thorough understanding of the synthesis – structure – property relationship. In this research, transition metal nitride and oxide thin films were deposited by atomic layer deposition (ALD), and the effect of deposition parameters on the structure and the chemical, physical, and functional properties were studied. ALD is a low temperature thin film deposition technique that involves sequentially exposing the substrate to two or more vapor-phase precursors, resulting in a self-limiting surface reaction. The self-limiting nature of the reaction enables strict control over the chemical composition, and the thickness in atomic scale. The nitride thin films, TixM1-xN (M = Mo, V, Hf) studied in this research for tribological applications were deposited using plasma enhanced atomic layer deposition (PEALD). ALD precursors for nitride deposition are insufficiently reactive and require high temperatures (~400 °C) to deposit high quality nitride films, at which it is difficult to maintain self-limiting growth conditions. Therefore, the plasma step in PEALD can provide reactive species that enable nitride formation at lower temperatures, as well as promote/introduce crystallinity even at deposition temperatures as low as 100 °C. Additionally, amorphous MoOx and HfO2 thin films were deposited by thermal evaporation and ALD respectively and studied for hole selective layers in photovoltaics and as gate dielectric respectively. Recently discovered PEALD TixV1-xN thin films resulted in the lowest wear rate of any ALD or nitride film explored to date. There can be a number of mechanisms contributing to excellent tribological performance, including grain boundary strengthening, inhibition of crack propagation by compressive residual stress, formation of lubricious tribofilm, and so on. However, the exact mechanisms resulting in the ultra-low wear rates of PEALD TixV1-xN thin film is not fully understood, because the bank of studies available in literature of ternary transition metal nitride for tribological applications by ALD are very scarce. Therefore, a wholistic research approach is required to understand the fundamental origins of record-level wear. In this research, the structure, and properties of the PEALD TixM1-xN (M = Mo, V, Hf) thin films were correlated to a series of deposition and processing parameters, including deposition temperature, substrate preparation method, applied substrate bias, the cation ratio in the film, and so on. In most cases the films were deposited at 250 °C at which the films were nanocrystalline with an FCC rock salt structure determined by grazing incident X-ray diffraction (GIXRD). The residual stresses of the films were also evaluated from GIXRD data using sin^2ψ method. X-ray reflectivity (XRR) was used to estimate the thickness and density of the films. The optical properties such as refractive index and extinction coefficient were determined from spectroscopic ellipsometry. The electrical resistivity of the films was evaluated using a four-point probe. The chemical composition of the films was determined using X-ray photoelectron spectroscopy (XPS). The wear rate and the coefficient of friction were determined from stripe test. Depositing TiVN at a series of temperatures showed that the films had best crystallinity at 250 °C deposition temperature resulting in the lowest wear rate at that temperature. It was found that the substrate cleaning method and adhesion layer also impacts the wear rate as well as the adhesion of the film to the substrate. TiMoN deposited on Si substrates cleaned by UV ozone exposure resulted in most adherent films compared to cleaning with organic solvents or the standard clean 1. Thicker SiO2 adhesion layer also resulted in more adherent films, but the wear rates were higher. TiN was deposited with tetrakis(dimethylamido) titanium (TDMAT) and N2 plasma. Hf (TDMAH), and Mo (bis(tert-butylimido)bis(dimethylamido) molybdenum) required N2-H2 plasma to deposit stoichiometric HfN and MoN with rock salt structure. However, some minor nitride and oxide phases were also formed during the deposition of HfN and MoN, but when alloyed with TiN, only a single-phase solid solution of TiHfN and TiMoN was formed that had rock salt structure. Increasing the plasma pulse from 10 s to 20 s for TiN deposition, saturates the plasma step and results in a densification from 88 % to 98 % of the theoretical density, as well as a reduction in electrical resistivity from 116 to 79 µΩ-cm. Application of RF substrate bias to up to 125 V caused the deposited TiMoN film to sputter away due to high energy plasma impingement. However, application of moderate bias, up to 50 V, caused densification and increase in compressive residual stress from 3.8 to 15.8 GPa, and as a result, the wear rate decreased from 5.4×10-7 to 3×10-8 mm3/Nm.
The amorphous MoOx and HfO2 on the other hand lack well-defined structure which makes it difficult to predict their properties. However, the short-range order of amorphous materials can be probed using X-ray absorption spectroscopy and X-ray total scattering, and the data were reduced to χ(R) and pair distribution functions, G(r) respectively. Melt-quench simulation by molecular dynamics (MD) was used to compute a starting amorphous structure for MoOx which was fitted against the experimental χ(R) and G(r) by Reverse Mote Carlo (RMC) simulation to get a realistic structural model. However, at the short-range order (< 5 Å), the HfO2 G(r) matches closely with the simulated G(r) of monoclinic HfO2, therefore monoclinic HfO2 was used as starting structure for RMC fitting. The refined structural models would be utilized to evaluate the electronic properties using density functional theory (DFT) calculations. This study is proof of concept for the experiment-theory symbiosis approach for amorphous materials. Correlations between atomic scale structural properties, electronic properties, and synthesis/processing conditions learned from such studies can be mined through the application of artificial neural networks (ANNs) providing the building blocks for data-driven materials design.</p
Schooling Performance and Formation Stability of Freely Swimming Bio-Robots
No full textThis thesis presents an experimental investigation into the schooling stability and performance characteristics of freely swimming bio-robots, focusing on various schooling configurations. A robust experimental setup was developed to facilitate the testing of two swimmers aligned in the stream-wise direction. Initial tests were conducted on each swimmer in isolation, examining one frequency across seven amplitudes. These tests provided a baseline for understanding individual swimmer dynamics. In the side-by-side schooling configuration, stability was observed in both in-phase and out-of-phase synchrony at specific lateral distances. This stability was primarily attributed to body-to-body interactions, with thinner tunas not exhibiting the same stability. Further experiments revealed that the swimmers also exhibited stability in staggered configurations. However, the interactions responsible for this stability were weaker compared to the side-by-side arrangement, dominated by body-to-body and body-to-tail stock interactions. In the in-line schooling configuration, no stability was detected. Across all tested configurations, no significant power benefits were observed under the basic conditions. Promising preliminary results were obtained in mismatched amplitude scenarios, indicating potential for power benefits. These findings suggest avenues for future research to optimize the power efficiency of schooling swimmers under varying conditions.</p
Assessing Coral Biodiversity Through Environmental DNA
No full textThe ocean provides humanity with many benefits including commercial, recreational, and subsistence fisheries; tourism economies; and natural resources and is a vital component of many human cultures. These benefits do not end in the surface ocean and coastal seas but extend to the ocean\u27s deepest depths and most remote locales. Diverse biological communities, including mesophotic and deep-sea coral communities, exist in the deep sea, provide ecosystem services, and are impacted by human activities. However, compared to shallow reefs, the fates of these deep coral communities in the face of these anthropogenic impacts receive comparatively less attention. Further, the biodiversity of deep coral communities and their distribution are less understood, largely due to the technical challenges in studying life at the depths that they occur. To rapidly assess and monitor coral biodiversity and its distribution, novel approaches are needed to complement conventional methods. One promising approach is the analysis of environmental DNA (eDNA), the DNA that all organisms release into their environments. However, to fully harness the power of this new technology, the spatiotemporal scales over which eDNA can be detected after it is shed from an organism into the ocean must be constrained.This dissertation research addresses this knowledge gap through a combination of experimental and field studies. Chapter 2 reports the results of eDNA persistence experiments conducted across combinations of temperatures, pH, and dissolved oxygen concentrations that reflect natural conditions in the subsurface open ocean. Coral eDNA degradation was accelerated at 20°C as compared to 10°C and 4°C, and degradation rates were weakly correlated with pH and not correlated with dissolved oxygen concentration. Acknowledging that temperature controls eDNA degradation rate, eDNA persistence times across the range of ocean temperatures can be approximated, informing the timescales over which eDNA may be transported in the field. In Chapter 3, bacterial community composition, assessed by sequencing the 16S gene, and bacterial abundance, measured with fluorescence microscopy, were measured from the experiments reported in Chapter 2. Experiments conducted at 20°C exhibited higher bacterial diversity, along with repeatable and faster rates of bacterial community turnover and bacterial growth than experiments conducted at lower temperatures. These results suggest that microbial processes, in synergy with temperature, govern eDNA degradation rate.In Chapter 4, new PCR primers and an accompanying reference sequence database for targeted eDNA metabarcoding of anthozoan corals at all depths are presented. These new PCR primers, which target the 28S gene, distinguish the vast majority of anthozoan corals at the genus level and enrich coral DNA from water samples. In Chapter 5, the transport of eDNA in the open ocean is practically investigated by sequencing eDNA at multiple depths and distances away from an offshore community of benthic invertebrates. Stratification in the water column resulted in limited detection of eDNA in the surface mixed layer and across depth zones in the thermocline. In sequencing data generated using the PCR primers for corals presented in Chapter 4, there was strong evidence for the horizontal transport of eDNA at distances up to ~1.5 kilometers from likely sources. This finding highlights the importance of interpreting species\u27 detections from eDNA data as a function of the sensitivity of the detection or sequencing protocol used. In whole, this dissertation provides new molecular tools and develops a conceptual framework for using eDNA analysis to conserve life in the ocean.</p