Search

In this thesis we address the problems associated to non-conjugate likelihood Gaussian process models, i.e., probabilistic models where the likelihood function and the Gaussian process priors are non-conjugate. Such problems include intractability, scalability, and poor local optima solutions for the parameters and hyper-parameters of the models. Particularly, in this thesis we address the aforementioned issues in the context of probabilistic models, where the likelihood's parameters are modelled as latent parameter functions drawn from correlated Gaussian processes. We study three ways to generate such latent parameter functions: 1. from a linear model of coregionalisation; 2. from convolution processes, i.e., a convolution integral between smoothing kernels and Gaussian process priors; and 3. using variational inducing kernels, an alternative form to generate the latent parameter functions through the convolution processes formalism, by using a double convolution integral. We borrow ideas from different variational optimisation mechanisms, that consist on introducing a variational (or exploratory) distribution over the model so as to build objective functions that: allow us to deal with intractability as well as enabling scalability when needing to hand massive amounts of data observations. Also, such variational optimisations mechanisms grant us to perform inference of the model hyper-parameters together with the posterior's parameters through a fully natural gradient optimisation scheme; a useful scheme for tackling the problem of poor local optima solutions. Such variational optimisation mechanisms have been broadly studied in the context of reinforcement and Bayesian deep learning showing to be successful exploratory-learning tools; nonetheless, they have not been much studied in the context of Gaussian process models, so we provide a study of their performance in said context.

This thesis is concerned with the understanding the contextual nature of consumer value creation within service encounters in high customer participation services. The study of service encounters is shown to have traditionally adopted individualistic or highly rational perspectives that ignore or overlook the significance of social context in shaping the material and cognitive resources that participants bring to the encounter and that they use to shape its outcomes and effectiveness. It is shown that by focusing on encounters as forms of social practice, involving a complex set of activities and skills, all of which need to be located within a context that reflects a wide range of factors (from individual biography to macrostructures) it is possible to develop a more complete understanding of how encounters are structured and how value is defined within them. This perspective provides a framework of concepts within which a detailed qualitative study of service encounters is conducted, remaining open to new possibilities that may be identified inductively. The study uses observation and interviews with participants in two case sites in Colombia, a higher education institution and a fitness club, to develop a framework that identifies issues that carry contextual relevance, summarizing these in a model of factors that need to be taken into account when investigating service encounters. Although the model is provisional, being derived from only two cases in a single country, it offers a depth of analysis that helps to identify the complex processes involved. It is suggested that further research will be needed to test this model for applicability in a broader context. In summary, the research contributes to a better understanding of service encounters by showing the importance of social context in shaping value-creation issues: participants of service encounters are not only decision makers but active human beings embedded in a cultural, socio-historical context, who construct value through service consumption practices that are dynamic, contextual and practical.

This thesis explores animic and perspectivist notions in the context of Nordic Seamanship with a biocultural framework. It examines the history, cosmologies, terminology, practices, physiology and phenomenology of Nordic crafts and arts of boat building, rope-making, seafaring and fishing. Rope-making, its molecular basis and the social organization in a boat reveal the way in which physical and social bodies coalesce in the harmonies of the differing intentionalities of their constituents, forming symmetric hierarchical structures, which are at the basis of Nordic egalitarian and individualistic society. Through the enskillment in seafaring and fishing, we explore the perspectival transformations involved in nausea; the development of sea-legs (the attunement to the rhythms of the sea), fishiness (empathy with the fish) and the meiths (a system navigation, perception and theorization of the coastal environment), showing the role of normal microbial biota in the perception and interactions with the environment. Based on the experience at sea, it is suggested that the ontologies developed through the interactions of seamanship constituted a cosmology that influenced the development of the Medieval Perspectivist theories in Natural Philosophy, Norse poetry and hermeneutics, which were means of secularization of pagan knowledge in the Nordic conversion to Christianity. Elaborating on some aspects of medieval perspectivist theory through their comparison with Amerindian animic theories and the biology of the eye it is suggested that its morphology entails an entoptic (inner-vision) microscopy, affording a means of visual perception and interaction with microbial entities. Finally, with the aid of a Treponema pallidum, a transatlantic traveller with a copious Amerindian mythology, it is shown that animic notions about spirits, dwarves and gods are coherent with an ecological physiology that takes into account microbial sociality and their role, both in health and in disease, in our metabolism, perception and relations with the environment in particular ecological communities. In so doing, it demonstrates that animic perspectivist ontologies are compatible with a naturalism that takes into account intentionality as a generalized physical property constituent of beings and things, and therefore sociality as generalized characteristic of the interactions between beings/things in the environment.

Maize (Zea mays ssp. mays), a crop of great importance worldwide, is characterised by its great adaptability and its extraordinary variability both at the molecular and the phenotypic level. Behind these remarkable features lies a complex evolutionary history. The main aim of this research is to contribute to a better understanding of it by answering questions about the origin, phylogenetic relationships, and initial dispersion of maize, using South America as a case study. To achieve this aim, archaeological specimens and primitive landraces were genotyped at 20 microsatellite loci. Sequences of Adh2-UM, a locus informative of the routes of initial dispersion of maize in South America, were also obtained. The data gathered were analysed using tree-building and network methods, and a Bayesian model-based clustering approach to infer population structure. Our results are consistent with a model of initial dispersion by which after its introduction, maize spread along the Andes and through eastern Bolivia and northwest Argentina towards the east of the continent to the area of Paraguay and southern Brazil, and gave rise to the maize distributed to the east of the Andean mountains. Analysis of population structure indicate that variation among the races analysed is more phenotypic than genetic, and that considerable amounts of gene flow has been occurring at a macro-regional level. We believe that cultural practices and cultural contacts played a fundamental role in making maize on these areas more genetically similar, while at the same time keeping maize phenotypically different. Further studies of ancient DNA, coupled with archaeological and ethnographical knowledge of cultural trajectories, have the huge potential of contributing to a better understanding of how and since when these macro-regional areas of gene flow have been in place.

The stress responses of cardiomyocytes are likely to constitute a significant aspect of the development of cardiac pathologies. Oxidative stress is a common theme in the pathophysiology of ischaemic and non-ischaemic cardiomyopathy. On the other hand, hypertrophic agonists (e.g. endothelin-1, ET-1) are important mediators of the hypertrophic response to hemodynamic overloading, and cardiac pathologies such as heart failure are usually preceded by cardiac hypertrophy. Microarray studies indicate that one of the genes most potently induced by H2O2 (as an oxidative stress) or ET-1 in cardiac myocytes is the transcription factor ATF3. This thesis examines the regulation and role of ATF3 in neonatal rat ventricular cardiomyocytes. The increase in expression of ATF3 mRNA was confirmed in neonatal rat cardiomyocytes exposed to H2O2 or ET-1 using RT-PCR and Q-PCR. Using immunoblotting, it was confirmed that H2O2 or ET-1 increased ATF3 protein expression. ATF3 is an immediate early gene since the upregulation of ATF3 mRNA by ET-1 was not inhibited by cycloheximide (20 μM). Upregulation of ATF3 was inhibited by U0126 (10 μM), suggesting that signalling through ERK1/2 (extracellular signal-regulated kinases 1/2) was required. Other studies suggest that ATF3 downregulates interleukin-6 (IL-6) in human cells. The rat IL-6 promoter possesses an ATF consensus sequence, and chromatin immunoprecipitation (ChIP) analysis indicated that ET-1 or H2O2 increased the association of ATF3 with the IL-6 promoter. IL-6 mRNA and protein expression was transiently upregulated in cardiomyocytes exposed to H2O2 or ET-1, and this was shown to be an immediate early gene response. Following IL-6 mRNA upregulation, its expression was more rapidly downregulated than that of ATF3. The peak of ATF3 protein expression coincided with the return of IL-6 mRNA to basal levels after stimulation with either ET-1 or H2O2. This suggests that ATF3 operates in a negative feedback loop to downregulate IL-6 mRNA expression in cardiomyocytes. This was confirmed by adenoviral-mediated overexpression of full-length ATF3 antisense RNA, which attenuated ATF3 mRNA and protein expression following stimulation with ET-1. Inhibition of ATF3 expression was associated with superinduction of IL-6 mRNA as shown by Q-PCR. Other potential downstream targets of ATF3 were identified from microarray studies using bioinformatics. Those with ATF/CRE consensus sequences in their promoters included epiregulin and leukaemia inhibitory factor (LIF) (also shown to be immediate early genes). Epiregulin and LIF mRNA expression was superinduced by ATF3 antisense RNA. Taken together, these experiments indicate that ATF3 operates in a negative feedback loop by downregulating a cluster of immediate early genes induced by ET-1. Thus, ATF3 may play an important role in the establishment and fine-tuning of an organised and compensatory hypertrophic response in the cardiomyocyte secondary to hypertrophic stimulation.

As student diversity in higher education continues to encompass myriad groups that include numerous intersecting combinations of backgrounds, higher education actors must be aware of the changing dynamics of the 21st century student. These changes include growing numbers of previously incarcerated and formerly gang-involved students seeking higher education as a way of reintegration. Faculty, staff, and students must learn about the so-called "normative" institutional practices that pose barriers for these college students, as well as discuss the harmful and demeaning effects that racial microaggressions can have. This critical race theory, funds of knowledge, and community cultural wealth study describes the experiences of incarcerated and formerly gang-involved students within higher education institutions, particularly focusing on community colleges. The findings present an alternative foundation for future empirical research and professional practices that serve these students. Primarily, critical race theory offers a counternarrative to seek inclusion of this otherwise marginalized student population. By acknowledging the embedded structural racist institutional practices along with individual enactments of racial microaggressions upon previously incarcerated and formerly gang-involved students within higher education, faculty, staff, and other students may better deal with the inclusion of other marginalized groups on campus, as well. Significantly, future empirical research on this student population requires direct attention to barriers that many higher education actors, policies, and practices cannot see, as they are particular to this group of students. For students seeking opportunities for educational advancement, the reality of a racialized social structure remains, often hindering their access and success within higher education systems (Harper, Patton, & Wooden, 2009). Addressing educational barriers faced by previously incarcerated, formerly gang-involved students requires a commitment to social justice and social equity. Researchers often occupy a unique position and ability to advocate for students by highlighting the lack of support, inclusion, and access to resources that address the needs of all students, including those in this group. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://www.proquest.com/en-US/products/dissertations/individuals.shtml.]

Manual inspection of telemetry data in the search for anomalies is a time-consuming threat detection technique. Most multi-signal systems send back extensive data that a single person cannot easily monitor in real time. Machine learning techniques that autonomously scan data and flag anomalies are attractive alternatives. The autonomous anomaly detection problem can be divided into two sub-problems: regression analysis and a classification process. In the regression analysis, a machine learning model is trained to reconstruct a given signal, and the classification process categorizes the reconstruction error as anomalous or nominal. This report examines the autonomous anomaly detection problem and proposes improvements to both the regression and classification sub-problems. With regard to the regression analysis, it was found that including the physics of the target signal in the machine learning model yielded a lower reconstruction error compared to a purely data-driven model. The classification approaches studied showed that cluster-based thresholding techniques accompanied by a pruning procedure can outperform non-parametric dynamic thresholds.

In 2001, The Institute of Medicine published its recommendations for bringing high quality care to all people of the United Sates. That solution involved fulfilling criteria expressed in the acronym, STEEEP. Care must be: Safe, Timely, Effective, Efficient, Equitable and Patient Centered (Institute of Medicine 2001). While improvements were made in terms of infant mortality, longevity, and deaths amenable to quality care, healthcare in the United States has remained fragmented with much work yet to be done. This leaves many uninsured individuals without access to affordable healthcare. Despite the implementation of policies such as the Affordable Care Act and the American Rescue Plan, which have expanded Medicaid and given access to many, it still falls short. Approximately 24.9 million people remain uninsured. The rising costs of healthcare in the U.S. have led to both insured and uninsured patients being exposed to medical debt, lower health status, and limited access to care. Safety net clinics, such as free clinics, have become essential for many uninsured individuals who rely on them to receive medical care. Free clinics are an example of safety nets that give medical access to the uninsured. These clinics have positive results on health outcomes and help to lower healthcare expenditures, particularly in emergency room visits. Studies have shown that uninsured individuals are more likely to use emergency services, which results in higher healthcare costs. Free clinics provide preventative care and early interventions that can help prevent costly emergency visits and hospitalizations. Moreover, free clinics serve as a place for volunteers to grow their skills and become better providers of medicine. Volunteers include physicians, nurses, medical students, and other healthcare professionals who dedicate their time and expertise to help those in need. Volunteers at free clinics are provided with a unique opportunity to enhance their skills by working with a diverse patient population that often has complex medical conditions. Free clinics are essential safety nets that provide medical access to the uninsured and underserved communities. Without these clinics, many uninsured individuals would be left without access to care, leading to poor health outcomes and higher healthcare costs. The importance of free clinics cannot be overstated, and unless there is a change in the current healthcare system, free clinics should be given the place they deserve, including more volunteer and funding support. As the U.S. healthcare system continues to evolve, it is critical to recognize the value of free clinics and the role they play in ensuring access to care for all individuals, regardless of their insurance status.

The physical and chemical functionality of solids is defined, in many aspects, by spatial inhomogeneity, with prominent examples in doped semiconductors or magnetic data storage. In materials with ferroic order, inhomogeneity manifests in the form of domains, that is, regions with different electric, magnetic, or structural long-range order separated by boundaries called domain walls. Magnetoelectric multiferroics are a particularly promising class of ordered materials because of their coexisting electric and magnetic orders rendering them candidates for low-energy or multi-state storage devices. In particular, the interplay between electric and magnetic order within domains and at domain walls endows magnetoelectric multiferroics with unique functional degrees of freedom. However, our understanding of these phenomena still resides at a fundamental level. This doctoral thesis unveils the electric and magnetic interplay on the domain level at the microscopic scale and discusses a control route to tailor the average domain size on a sub-micron length scale. In addition, novel methodologies to probe domain populations below the optical resolution limit are explored. As a model compound, hexagonal manganites are investigated. In these compounds, electric and magnetic orders arise independently, and therefore the interplay between the orders is expected to be weak or non-mandatory. This is in contrast to multiferroic materials where the electric order is directly induced by the magnetic order, and the interplay is therefore inherently pronounced, leading to one-to-one rigidly coupled magnetoelectric domains. Importantly, however, the independence of electric and magnetic orders in the hexagonal manganites can leave the freedom for a richer landscape of domains and domain walls. It is shown in this doctoral thesis that hexagonal manganites indeed show a rich variety of domain walls. More specifically, three types of magnetic(-electric) domain walls are identified, along with magnetic vortices that form between the different domain wall intersections. This rich landscape of domain walls and vortices has its origins --- contrary to general belief --- in a pronounced bulk magnetoelectric coupling phenomenon present in these materials. Moving forward, another inherent aspect that determines the functionality of ordered materials is the average domain size which, in principle, defines how small an information ``bit'' can be made. This doctoral thesis explores chemical substitution as a route to tailor the domain size in hexagonal manganites from the micron- down to the angstrom range while preserving the inherent bulk magnetoelectric coupling mechanism. The investigation of domains in the sub-micrometer range requires imaging techniques that overcome the limitations of optical resolution. To circumvent this limitation, an experimental proof-of-concept is presented. More specifically, by combining pixel-by-pixel analysis and numerical simulations, we expand the capabilities of far-field optical second-harmonic generation (SHG) to probe nanodomain populations. In addition, combined SHG and scanning probe techniques (SPM) enable studies on the level of individual domain walls. The research results presented in this doctoral thesis, hence, provide insights that expand the field of multiferroics from a fundamental but also from a methodical point of view. These findings are relevant not only for bulk multiferroic systems but could be extended to the emergent field of thin-film multiferroics, for example to explore magnetoelectric coupling effects in hexagonal manganites thin films or to tailor magnetoelectric domains during synthesis. These results highlight the importance of microscopic inhomogeneity, specifically domains and domain walls, and its potential to promote their magnetoelectric functionality in these materials.