22689 research outputs found
Sort by
Physical Properties and Chemical Reactivity of Size-Selected Atoms, Molecules, and Clusters: Gas Phase and Surface Studies
Get PDFAnion photoelectron spectroscopy gives insight into the electronic structure of the probed atom, molecule, or cluster anion. This was used to study the physical properties as well as the chemical reactivity of a wide range of systems. One advantage of experiments on mass-selected species in the gas phase is that no solvent effects, etc., need to be considered. They give a tool to measure properties that can also be calculated and evaluate theoretical models.
These methods were applied to a wide range of different systems including,
Uranium compounds (Chapter II), metal hydride clusters for CO2 hydrogenation (Chapter III), acid and base dimers (Chapter IV), and Iridium catalyzed Monopropellants (Chapter V).
A new method was implemented for the Time-of-flight mass spectrometer to investigate the effects of Penning detachment on SF5. Anions (Chapter VI).
Additionally, to the gas phase studies, surface studies were conducted for the reactivity of ionic liquids (Monopropellants) decomposition on Iridium supported by HOPG. XPS and TPD/R were utilized. The Sarine gas simulant DMMP decomposition catalyzed by different mixed metal oxide clusters supported by HOPG was also investigated with those methods (Chapter VII).
In the appendix, supporting information for the used machines can be found
Label-free morpho-molecular cellular analysis with Raman and quantitative phase imaging
Get PDFImaging is a fundamental source of information and is used across medical and scientific disciplines to visualize, assess, and monitor the presence, status, and progression of various pathological and physiological conditions. Notably, significant progress in imaging technology has occurred, involving technical enhancements in main components, integration of multiple techniques, and improved processing capability, and thereby facilitated providing information in a more timely, accurate, and reliable manner. In light of these breakthroughs, we worked on the imaging of cells and tissues through various label-free optical modalities, including Raman spectroscopy and optical diffraction tomography.
In this thesis, we present three case studies in which we characterized the molecular composition and morphology traits unique to renal amyloidosis, cancer senescence, and neurodegeneration for cell and tissue diagnostics. These studies led to the development of a novel workflow for informative imaging and artificial intelligence (AI)-assisted identification of cell morphology and physiology without exogeneous labels. Through these works, we aimed to better understand biological processes and pathological conditions in order to further develop current diagnostic pipelines.
First, we studied the two most common protein subtypes of renal amyloidosis: light chain amyloid and serum amyloid A. We characterized the molecular features unique to each subtype and used these to distinguish them with a combination of Raman spectroscopy and machine learning algorithms for automated tissue diagnosis. Second, we monitored and quantified the chemical, morphological, and physiological changes linked to cancer senescence through multiple complementary optical techniques, including coherent Raman scattering, multi-photon absorption, and optical interferometry. We discovered multiple early-to-late therapy-induced senescence manifestations of human liver carcinoma cells, including mitochondrial rearrangement, an increase of volume and dry mass, and the accumulation of lipid vesicles. Lastly, we developed a multimodal imaging workflow enabling both Raman spectroscopic and matrix-assisted laser desorption/ionization mass spectrometry imaging (MALDI MSI). We demonstrated the successful imaging of mouse liver homogenate, kidney, and brain tissue through a one-step sample preparation with a single sample without compromising spectral and spatial quality
Spinoza's Metaphysics of Infinity
No full textGiven the abundance and prominence of infinities in Spinoza's edifice, the limited amount of literature on the topic might surprise. God, attributes, modes perfection, power, reality, motion and
rest, the inner striving of things (conatus) are either defined through infinity or can at least have some infinite instantition. Furthermore, from Leibniz through Kant, Hegel, Cantor up to Deleuze, Spinoza’s take on infinity has been a crucial touchpoint of philosophical controversy. Yet commentators have restricted themselves to scattered remarks and partial analyses. In the following pages, I attempt to close that gap with what I call a deflationary account of Spinoza’s infinities. The deflationary account reads as follows: absolute infinity (God) and infinity in its own kind (attributes) and infinity in modes (such as infinitely many rectangles that may be conceived within a circle) work in the exact same way: they all follow from an indetermined essence. This account is
deflationary because it offers one theory that captures infinity across all three ontological states (substance, attributes, modes). Basically, my point is that we can understand what Spinoza means by infinity if we look at his theory of simple geometric objects (circles, rectangles) first. Or inversely, what I want to say is: God is really a geometric object like a circle – just a little more complicated
Design and Synthesis of a Novel Class of RNA Polymerase I Inhibitors
Get PDFRNA Polymerase I (Pol I) is one of three DNA-dependent RNA polymerases and is responsible for transcription of the 47S ribosomal RNA (rRNA) precursor. The 47S rRNA transcript is subsequently processed to release the 18S, 5.8S, and 28S rRNAs which are assembled into ribosomes. Pol I transcription serves as the rate-limiting step in ribosome biogenesis, accounting for up to 60% of active transcription in eukaryotic cells, directly influencing protein accumulation, cell growth, and cell division. Many types of cancers exhibit dysregulated rates of Pol I transcription, reflecting a need for increased ribosome synthesis to generate proteins to sustain heightened growth rates. Cancer cells may be selectively vulnerable to agents that inhibit Pol I transcription, providing an attractive therapeutic strategy for cancer treatment. BMH-21 is the first specific and selective Pol I inhibitor and does so by intercalating into GC-rich rDNA and creating a transcription block, leading to the ubiquitination and proteasomal degradation of the large catalytic subunit, RPA194. Notably, it accomplishes this independently of p53 and without eliciting a DNA damage response. BMH-21 is the first of only a small number of compounds to exhibit the RPA194 degradation phenotype, and a quantitative cell-based assay has been developed to measure the extent of RPA194 degradation caused by compound treatment. The primary goal of this work is to design and synthesize small molecule inhibitors of Pol I, with a focus on determining key pharmacophores and generating structure-activity relationship (SAR) data. SAR studies revealed key pharmacophores, but also showed that activity was limited within narrow chemical space. Further SAR efforts, summarized by this work, have produced additional scaffolds as well as addressed some off-target activity while maintaining desired RPA194 degradation potency. In collaboration with Evotec, the RPA194 degradation assay was translated from 96-well plate format to 384-well plate format to facilitate high-throughput screening (HTS) efforts, providing the opportunity to discover new Pol I inhibitors and to generate new SARs. Efforts were made to transform BMH-21 into a chemical probe to gain structural insight about its binding interactions and to identify its molecular target, providing rationale for the improved design of future compounds
Improving In-context Machine Translation
Get PDFIn-context Learning for Machine Translation (MT) is a new paradigm enabled by very large language models self-supervised on massive amounts of language data. These models have demonstrated potential to translate between multiple languages using monolingual data. While they still lag behind supervised systems, the gap appears to be closing.
In the first part of the thesis, we conduct early investigations into improving In-context Machine Translation, keeping consistent with the unsupervised (monolingual) promise of this paradigm. The methods we develop are lightweight in both computation and data resource such that regular practitioners can easily adopt them to improve In-context MT systems. We introduce Prefix Embeddings which only requires monolingual data to train, and induces the model to generate in the target language. Having identified prefixes as an important aspect of inducing a model to translate in the target language, we propose an automatic method of initialising this prefix instead of an arbitrary manual one. Next we consider an alternative way to improve translation at the output stage via the decoding step. We propose an Anti-Language Model decoding objective with decay time-step for reducing the influence of the dominant source language.
The second part of the thesis makes contributions towards understanding In-context Machine Translation. We challenge the dominant notion that good prompts must be lexically and semantically similar to the test sentence. Instead, we suggest that coherence of the prompt examples and the test sentence are more important for performance, and demonstrate this using domain and document level experiments. The findings of this work have practical implications for both style adaptation and Document Level Machine Translation. Finally, we go beyond the input/output of the model in an attempt to understand In-context MT at a deeper level. Through a series of layer-wise masking, positional-masking and compression analysis, we locate where instructions and examples are processed in the model, and find evidence that the distribution of MT attention heads fundamentally differs from supervised Neural Machine Translation
Fostering Faculty Engagement and Knowledge Sharing in Higher Education
Get PDFThe increasing reliance on adjunct faculty in higher education calls for innovative strategies to address their professional development needs and foster a sense of community among educators. This dissertation study focuses on an independent art college, which confronts the unique challenge of a fully adjunct faculty and a fragmented urban campus with limited shared spaces. This situation has led to historically low faculty participation in professional development activities and a reported lack of community among instructors. Professional development research shows that ongoing interaction and collaboration among faculty members is required to improve teachers’ classroom practice. So, the college’s pending mass retirement within its experienced faculty population heightens the urgency of cultivating a community of shared learning. The findings from in-depth interviews with faculty members in a needs assessment study revealed a faculty culture characterized by isolation and a desire for more meaningful interactions with their teaching peers. Drawing from insights gained through synthesizing literature on self-determination theory (SDT) and participatory design research, the proposed treatment is the design and implementation of a virtual faculty common (FC). The treatment process utilizes the participatory design approach to drive iterative development and to promote the FC from an authentic faculty perspective. Principles from SDT are incorporated into the FC design and communication strategy to support sustained use of the FC and to inform the methodology for the subsequent impact study. The faculty common impact study employs a mixed methods research design to assess the effects of increased knowledge sharing and collaboration with basic psychological needs satisfaction as the mediating variable
Men's use of intimate partner violence in armed conflict: Evidence for intervention
No full textIntimate partner violence against women (IPVAW) is a critical determinant of mental, physical, and reproductive health for women in conflict-affected settings. Although attention to IPVAW in conflict-affected settings is growing, it remains incompletely understood and insufficiently addressed. Interventions engaging men have mixed evidence of efficacy. The objective of this dissertation is to provide evidence that can support interventions engaging men to reduce IPVAW in conflict-affected settings. In chapter one, we introduce the problem of IPVAW in conflict-affected settings and detail the specific aims of this dissertation. In chapter two, we review the literature on concepts related to these specific aims, which elucidate who has elevated risk for using IPVAW, describe which theory-informed pathways predict IPVAW use, and synthesize evidence pointing to trauma as a focus for IPVAW intervention in conflict- and violence-affected settings. Chapter three uses a classification tree analysis of men in Eastern Democratic Republic of Congo (DRC) to characterize men with higher risk of using IPVAW. Our findings indicate the importance of three attitudes and experiences ¬¬– gender inequitable attitudes, trauma, and human insecurity – highlighting the need to identify the complex realities of men as victims and potential perpetrators of violence. In chapter four, we perform structural equation modeling, using the same data from Eastern DRC to test pathways suggested by four theories that describe how risk factors increase past-year use of IPVAW in conflict-affected settings. Our findings partially support feminist and mental health theories, suggesting opportunities to add trauma-focused care to current gender attitudes/norms programming to reduce the burden of IPVAW in conflict-affected settings. Chapter five synthesizes evidence on the relationship between trauma and IPVAW, discussing empirical and theoretical links, intervention, and recommendations. We argue that addressing trauma symptoms represents a key, although neglected, opportunity to reduce IPVAW in conflict- and violence-affected settings. Finally, in chapter six, we discuss the implications of these findings for research and practice. The targeting and approaches that this dissertation illuminate have the potential to improve men’s mental wellbeing, reduce the burden of IPVAW, and consequently improve women’s mental wellbeing in settings of armed conflict and collective violence globally
Preparedness and opinions of nurses towards intimate partner violence in China: A mixed methods study
Get PDFBackground: Intimate partner violence (IPV) is a serious global health issue. The purpose of the dissertation is to describe characteristics of Chinese nurses in relation to perceived preparedness to manage IPV and opinions on IPV, and to explore the barriers and facilitators in responding to IPV by nurses in China.
Methods: The study used an explanatory sequential mixed-methods design. Two scales from the Physician Readiness to Manage Intimate Partner Violence Survey (PREMIS) were adapted and validated. From September 3 to 23, 2020, an online survey was conducted in two tertiary general hospitals in China. From September 15 to December 23, 2020, 43 semi-structured interviews were conducted.
Results: The Chinese versions of Perceived Preparation and Opinions scales demonstrated acceptable psychometric properties. In total, 1071 survey respondents (1039 female [97.0%]) and 43 interview participants (34 female [79.1%]) were included. The survey respondents largely agreed with feeling prepared to manage IPV. The findings of surveyed opinions (i.e., Response competencies, Routine practice, Actual activities, Professionals, Victims, and Alcohol/drugs) were mixed and intertwined with social desirability bias. The quantitative and qualitative data were consistent, contradicted, and supplemented. Totally 29 barriers and 14 facilitators were identified.
Conclusions: This study is the first to adapt and evaluate the psychometric properties of the Perceived Preparation and Opinions scales from the PREMIS for use among Chinese nurses. This study also is the first to describe the current status of nursing practice in relation to IPV prevention and intervention in China. There was a lack of actual preparedness to manage IPV among nurses in China, and their opinions on IPV were mixed and included socially desirable responses. Commonly reported and previously unreported barriers and facilitators in responding to IPV were identified. The Confirmation, Discordance, and Expansion findings underscore the importance of a mixed-methods design. The results of this study support World Health Organization recommendations for selective screening, highlighting the great potential of nurses to respond to IPV in low- and middle-income countries where recognition is limited, policies are lacking, resources are scarce, and education and training is generally absent
AT THE THRESHOLD OF POSSIBILITY: CULTIVATING ECOLOGICAL RELATIONALITY IN NEOLIBERAL TURKEY
No full textSince the early 2000s, an increasing number of secular white-collar professionals in urban areas in Turkey have been turning to sustainable ways of living, consuming food grown without synthetic inputs, organizing in alternative food networks, and moving to the countryside to take up ecological farming as a livelihood – all with the aim of creating alternatives to the conventional agri-food system. Placing this phenomenon within the social and political history of Turkey, I seek to understand how this back-to-the-land movement came into being, and what kinds of ethical labor people engage in to sustain it within a neoliberal agricultural economy and under conditions of political authoritarianism. By simultaneously focusing on daily consumption habits as acts of self-cultivation and various forms of collective action, I ask: how do participants’ imaginations of the good life intersect with and at times confound their visions of action and social change for an alternative agroecological future? How do these urbanites negotiate between their aspirations for larger transformation and the various constraints they face—personal, interpersonal, organizational, economic, and political?
The research relies on 18 months of ethnographic fieldwork conducted in urban and rural settings with back-to-the-landers, including food activists, ecological living enthusiasts, and farmers, crosscutting a variety of spaces, including consumer and producer cooperatives, meetings of organized groups, farms, open-air markets, and people’s homes and kitchens. Combining participant observation, semi-structured interviews, and discourse analysis of texts circulating in these networks, I conceptualize the notion of “back-to-the-land” not simply as a movement across space, but as a countercultural ethos that involves developing transformative connections between the country and the city. Instead of rushing to practice theory or the framework of neoliberal governmentality to study the inevitable tension between individualized lifestyle politics and collective action as it emerges among back-to-the-landers, I suggest turning to the anthropology of ethics and scrutinize the vibrant moral life that emerges at the interstices of structure and agency, possibility and impossibility, and intention and effect. By taking seriously people’s own theorization of social change, such study, in turn, provides insights into what it means for subjects with privilege to perform moral striving at individual and collective levels to enact change, and subsequently deal with the ineffectiveness of their actions
Building and Evaluating Open-Vocabulary Language Models
Get PDFLanguage models have always been a fundamental NLP tool and application. This thesis focuses on open-vocabulary language models, i.e., models that can deal with novel and unknown words at runtime. We will propose both new ways to construct such models as well as use such models in cross-linguistic evaluations to answer questions of difficulty and language-specificity in modern NLP tools.
We start by surveying linguistic background as well as past and present NLP approaches to tokenization and open-vocabulary language modeling (Mielke et al., 2021).
Thus equipped, we establish desirable principles for such models, both from an engineering mindset as well as a linguistic one and hypothesize a model based on the marriage of neural language modeling and Bayesian nonparametrics to handle a truly infinite vocabulary, boasting attractive theoretical properties and mathematical soundness, but presenting practical implementation difficulties.
As a compromise, we thus introduce a word-based two-level language model that still has many desirable characteristics while being highly feasible to run (Mielke and Eisner, 2019). Unlike the more dominant approaches of characters or subword units as one-layer tokenization it uses words; its key feature is the ability to generate novel words in context and in isolation.
Moving on to evaluation, we ask: how do such models deal with the wide variety of languages of the world---are they struggling with some languages? Relating this question to a more linguistic one, are some languages inherently more difficult to deal with?
Using simple methods, we show that indeed they are, starting with a small pilot study that suggests typological predictors of difficulty (Cotterell et al., 2018). Thus encouraged, we design a far bigger study with more powerful methodology, a principled and highly feasible evaluation and comparison scheme based again on multi-text likelihood (Mielke et al., 2019). This larger study shows that the earlier conclusion of typological predictors is difficult to substantiate, but also offers a new insight on the complexity of Translationese.
Following that theme, we end by extending this scheme to machine translation models to answer questions traditional evaluation metrics like BLEU cannot (Bugliarello et al., 2020)