136879 research outputs found
Sort by
Cultural Pathways
No full textThis project allows you to experience and meditate on the journey. Any journey that has a beginning eventually comes to an end, but we always have the present moment, and sometimes a chance to reflect on the past. The beginning of the journey is represented by the buildings approach, the end is represented by the museum���s final meditation space, and the present moment in time is represented by the central area of the site. Along the way you get the opportunity to see Nepal���s past represented in their rich and colorful artifacts. However, unlike many other museums, these artifacts sit in front of the breathtaking Himalayas which have integrated with Nepal���s culture for hundreds of years. It is the goal of Cultural Pathways, that the user is able to meditate on the journey, the nature, Nepal���s rich past, and its beautiful present
Leveraging Economic Interdependence for Offshore Energy Developments
Get PDFWhy is the certification of potential energy reserves in the Levant Basin Province (LBP) lagging so far behind that of its oil- and natural-gas-rich neighbors? The answer lies in the geopolitically fragmented nature of the LBP. Unlike the LBP, Egypt���s Eastern Mediterranean resources fall neatly within its Exclusive Economic Zone (EEZ). The political fragmentation of a resource basin that cuts across often-disputed maritime borders, however, constitutes a major constraint on its development.
This paper examines how geopolitical fragmentation negatively impacts the development of offshore energy resources in the LBP. By reviewing recent events in the region, it underscores the complications associated with maritime border disputes and the exploration, extraction, and transportation of cross-border resources. The paper then draws attention to how adjacent regions have navigated similar challenges by capitalizing on interdependent economic opportunities.Mosbacher Institute for Trade, Economics, and Public Policy
Bush School of Government & Public Servic
Advancing Geometric Deep Learning: From Fundamental Approaches to Drug Discovery Applications
No full textGeometric deep learning (GDL) integrates deep learning techniques with geometrically structured data such as graphs, 3D geometries, and point clouds. In this dissertation, we particularly consider graph and geometric graph data and aim to contribute to the advancement of geometric deep learning by developing novel fundamental approaches and exploring drug discovery applications.
Our primary methodological contribution focuses on graph neural networks (GNNs). Common GNNs, using message passing, suffer from performance issues when aiming for deeper architectures. Several prior studies attribute such performance deterioration to the over-smoothing issue, where repeated propagation makes node representations of different classes indistinguishable. Our analysis identifies the entanglement of representation transformation and propagation as the key issue. By decoupling these operations, we enable deeper GNNs to learn from larger receptive fields without performance deterioration. We introduce the Deep Adaptive Graph Neural Net-work (DAGNN) to adaptively integrate large receptive field information. Experiments on various datasets confirm the benefits of our approach.
Next, we address the expressiveness limitation of GNNs, which are restricted by the 1-dimensional Weisfeiler-Leman (1-WL) algorithm. We propose a scalable and powerful GNN framework, NC-GNN, which leverages a neural adaptation of an enhanced isomorphism test, NC-1-WL. The ex-pressiveness of NC-1-WL surpasses 1-WL, and experiments show the efficacy of our NC-GNN on benchmarks.
Furthermore, while graph representation learning has been extensively studied, generative modeling on discrete graph data remains challenging. We introduce a principled approach using energy-based models (EBMs). Traditional EBMs face difficulties with discrete data, so we propose a method, ratio matching with gradient-guided importance sampling (RMwGGIS), which harnesses energy function gradients for more efficient learning. Our experiments demonstrate its effectiveness and scalability.
Lastly, we venture into the generation of 3D geometric graphs, which is crucial for drug discovery. We introduce GraphBP, which generates 3D molecules for target protein binding. Experiments demonstrate that our GraphBP is effective in generating 3D molecules with a high binding ability to target proteins
Structural and Morphological Investigation of Carbon Based Materials for Electrochemical Applications
Get PDFBio-derived carbon-based-supercapacitor electrodes present an economical solution as sustainable sources for energy storage. Serious knowledge gaps exist for bio-derived carbon electrodes in terms of causes of largely unknown surface chemistry, poor volumetric energy density, and uneven pore structure and distribution which adversely affect electrochemical performance, cost, and scalability. To date, it is not yet clear how the electrochemical performance is affected by chemical composition and molecular structure in bio-derived carbon. In addition, the relation of surface morphology and microstructure with electrochemistry has not been established. This research attempts to fill those knowledge gaps with new understanding of the effects of bio-derived carbon structure and morphology on electrochemical performance.
The goal of this research is to obtain new knowledge in terms of biomass electrochemistry to design and engineer sustainable and high-performance electronics. To reach the goal, this research will investigate the structural and morphological characteristics of lignin based materials and their effects on electrochemical performance as electrodes. Specifically, it is intended to understand 1) how the presence of different chemical structures and 2) particle morphologies of lignin affect the charge storage and cycle life of a quasi solid-state supercapacitor. To carry out the research, experimental approaches combined with theoretical analysis will be conducted. Specific tasks include fabrication, assembly, test, analysis, and optimization of supercapacitor electrodes using various bio-derived lignin differing in structural identity and morphological features.
Lignin varying in structural identity (alkaline lignin (AL), lignosulfonate (LS), and dealkaline lignin (DAL)) and morphology (micro-fiber lignin and nano-spherical lignin) are used to evaluate electrochemical behavior. Specific capacitance, retention, impedance, energy, and power density results are compared and analyzed.
This research generates new knowledge in understanding lignin electrochemistry that will aid in significant advances in the future design of bio-derived carbon based electrochemical devices. Novel synthesis processes for electroactive lignin micro and nanoparticles will be developed. Unique design of lignin supercapacitors will be accomplished and improved to attain high performance.
This thesis contains seven chapters. Following the background introduction in Chapter I, the motivation and objectives of this work are presented in Chapter II. This is followed by the research on structural investigation of micro-lignin fibers: electrochemical behavior of alkali lignin fibers (Chapter III), transition metal oxide (NiWO���) nanoparticles doped lignin (Chapter IV), and effects of hydrothermally impregnated MnO��� particles on lignin fibers (Chapter V). Structural investigation on nano-spherical lignin particles is covered in Chapter VI. Finally, a summary and commentary on future directions of research are provided in Chapter VII
Microfluidic Systems for Interrogating Host-Pathogen Interactions
Get PDFMicroorganisms in the environment is extremely diverse, yet those that are identified and their functions known are extremely small. Humans have been increasingly coming in contact with these diverse microorganisms stemming from increases in global transportation and urbanization. The emergence of readily available gene editing tools as well as continuing evolution of these microorganisms as they come in contact with human make the biocomplexity even higher. Together, the potential bacterial pathogens in these environments pose significant risk to the public. Methods and tools that can provide comprehensive, systematic, and rapid analyses of these microorganisms can greatly contribute to our understanding of these extremely complex microbial world, especially for microbial pathogens. The research presented here provides three new strategies: first, a proactive strategy to study and understand the molecular mechanisms by which pathogens emerge and evolve, and second, an effective surveillance system that is capable of detecting and analyzing the emerging pathogens in timely manner, and third, development of microfluidic technologies that can enable high-throughput investigation of biological samples. With these strategies in mind, this dissertation presents the development, testing, and utilization of several novel microfluidics systems that each allow different microbial interrogation approaches to be performed in a high-throughput lab-on-a-chip format for combating the emergence of microbial pathogens.
The first platform developed is a microfluidic system named SEER platform, System for Evaluating the Emergence of Replicating pathogens, which enables the fully automated, multi-round directed evolution of intracellular parasitism in the laboratory. The SEER platform utilizes a porous membrane filter-based selective cell manipulation microfluidic technology and can direct na��ve bacterial populations that are initially incapable of intracellular bacterial parasitism to evolve and generate populations that can display enhanced survival within macrophages, so to determine genetic loci that confer this phenotype. This platform was successfully utilized to study the symbiotic evolutionary process between E. coli DH5�� strain and RAW264.7 macrophage and have confirmed the contribution of cpxR gene on the enhanced survival phenotype. The second platform is a high-throughput, dielectrophoresis-based microfluidics platform that can achieve selective manipulation of cells from mixed cell communities in a non-destructive, single-cell resolution manner, which allows microorganisms that adhere to mammalian host cells to be selected and sorted for further analysis of their pathogenicity. This platform was successfully utilized to investigate two environmental soil samples and various adherent pathogens that originally presented in the soil with low abundancy were extracted and identified with this high-throughput microfluidic method.
In addition to the successful development of two microfluidic systems for host-pathogen interaction studies, several microfluidic technology advancements have also been achieved based on the utilization of dielectrophoresis phenomena. These include the development of in-droplet cell separation technology, in-droplet solution exchange technology, droplet size-based sorting technology, as well as new microfabrication architectures that can significantly improve the performances of microfluidic systems. All developed technologies have been successfully validated and their utilities demonstrated and are expected to greatly expand the potential application of microfluidic systems in conducting cell biology assays
