Alliance One Tobacco (Malawi)
Academic Research Repository at the Institute of Developing EconomiesNot a member yet
52305 research outputs found
Sort by
Setting Gendered Expectations? Recruiter Outreach Bias in Online Tech Training Programs
No full textCompetence development in digital technologies, analytics, and artificial intelligence is increasingly important to all types of organizations and their workforce. Universities and corporations are investing heavily in developing training programs, at all tenure levels, to meet the new skills needs. However, there is a risk that the new set of lucrative opportunities for employees in these tech-heavy fields will be biased against diverse demographic groups like women. Although much research has examined the experiences of women in science, technology, engineering, and mathematics (STEM) fields and occupations, less understood is the extent to which gender stereotypes influence recruiters’ perceptions and evaluations of individuals who are deciding whether to apply to STEM training programs. These behaviors are typically unobserved because they occur prior to the application interface. We address this question by investigating recruiters’ initial outreach decisions to more than 166,000 prospective students who have expressed interest in applying to a midcareer level online tech training program in business analytics. Using data on the recruiters’ communications, our results indicate that recruiters are less likely to initiate contact with female than male prospects and search for additional signals of quality from female prospects before contacting them. We also find evidence that recruiters are more likely to base initial outreach activities on prospect gender when they have higher workloads and limited attention. We conclude with a discussion of the implications of this research for our understanding of how screening and selection decisions prior to the application interface may undermine organizational efforts to achieve gender equality and diversity as well as the potential for demand-side interventions to mitigate these gender disparities.Version of Recor
Large Scale Inference and Combinatorial Variable Selection for Complex Dataset
No full textThis dissertation advances the field of modern statistical theory and methodology by focusing on two primary areas: first, the quantification of uncertainty beyond mere estimation in combinatorial inference theory; and second, addressing the complexities and challenges inherent in electronic health records (EHR).
Chapter 1 introduces a novel combinatorial inference framework to conduct general uncertainty quantification in ranking problems. By considering the Bradley-Terry-Luce model, we aim to infer both local and global ranking properties, and generalize the method to multi-tesing problem with false discovery rate (FDR) control.
Chapter 2 focuses on the development of a semi-supervised approach that efficiently leverages sizable unlabeled samples with error-prone EHR surrogate outcomes from multiple local sites, to improve the learning accuracy of the small gold-labeled data. we apply our method to develop a high dimensional genetic risk model for type II diabetes using large-scale data sets from UK and Mass General Brigham biobanks, where only a small fraction of subjects in one site has been labeled via chart reviewing.
Chapter 3 presents
a novel inferential framework for general graphical models to select graph features with false discovery rate controlled.
The proposed method is based on the maximum of -values from single edges that comprise the topological feature of interest, thus is able to detect weak signals.
Moreover, we introduce the -dimensional persistent Homology Adaptive selectioN (KHAN) algorithm to select all the homological features within dimensions with the uniform control of the false discovery rate over continuous filtration levels. The KHAN method applies a novel discrete Gram-Schmidt algorithm to select statistically significant generators from the homology group.
We apply the structural screening method to identify the important residues of the SARS-CoV-2 spike protein during the binding process to the ACE2 receptors. We score the residues for all domains in the spike protein by the -value weighted filtration level in the network persistent homology for the closed, partially open, and open states and identify the residues crucial for protein conformational changes and thus being potential targets for inhibition
Revivals imply quantum many-body scars
No full textWe derive general results relating revivals in the dynamics of quantum many-body systems to the entanglement properties of energy eigenstates. For a D-dimensional lattice system of N sites initialized in a low-entangled and short-range correlated state, our results show that a perfect revival of the state after a time at most poly(N) implies the existence of "quantum many-body scars", whose number grows at least as the square root of N up to poly-logarithmic factors. These are energy eigenstates with energies placed in an equally-spaced ladder and with Rényi entanglement entropy scaling as log(N) plus an area law term for any region of the lattice. This shows that quantum many-body scars are a necessary condition for revivals, independent of particularities of the Hamiltonian leading to them. We also present results for approximate revivals, for revivals of expectation values of observables and prove that the duration of revivals of states has to become vanishingly short with increasing system size.Accepted Manuscrip
Trioedd Ynys Prydain and the Transmission of Medieval Welsh Narratives
No full textABSTRACT
Dissertation Advisor: Catherine McKenna Author: Celeste L. Andrews
Trioedd Ynys Prydain and the Transmission of Medieval Welsh Narratives
Trioedd Ynys Prydain are a series of closely related texts, appearing in manuscripts from the second half of the thirteenth century forward, which serve as compendia of cyfarwyddyd (storytelling material) organized in triads, or groupings of three. Texts of these Triads survive in nineteen manuscripts dating from the thirteenth to the sixteenth centuries. This dissertation examines and contextualizes those nineteen texts to better understand how the cyfarwyddyd contained within them developed and circulated over that three-century period.
It approaches this effort in two ways. The first is to closely examine individual triads to identify their sources and better understand their transmission history. Part 1 consists of case studies of six triads, all of which belong to the category of “expanded narrative triads,” those triads which resemble narrative prose passages. I argue in these case studies that these triads should be considered authored texts in their own right, and that each reflects the creative choices of its individual author.
The second approach, in Part 2 of the dissertation, is to think about how each surviving text of Trioedd Ynys Prydain came to be. I argue that the texts can often be shown as collections of multiple shorter, earlier texts which have been compiled by scribes and copyists. I have called these “microtexts.” For the purposes of this dissertation, a “microtext” is a sub-section within a text of Trioedd Ynys Prydain which likely has its own transmission history and should be considered its own discrete part of the larger text. This dissertation argues that close readings of three texts of Trioedd Ynys Prydain – those in Peniarth 47, the Red Book of Hergest, and Peniarth 50 -- provide convincing evidence that the scribes who developed these texts did so by compiling several shorter collections of Trioedd Ynys Prydain together. These shorter collections which were compiled together are microtexts.
Taken together, these two approaches demonstrate that Trioedd Ynys Prydain are authored texts that have been composed, expanded, and re-worked by the individuals who worked with this material over a three-century period between the thirteenth and sixteenth
Examining the role of succinate signaling in tissue remodeling
No full textWhile the benefits of exercise on systemic health are undeniable, vigorous muscle contraction, like the ones associated with exercise, puts considerable mechanical and energetic strain on muscle fibers. To meet these newfound demands, skeletal muscle must undergo extensive remodeling. This is a highly coordinated process that involves the participation of numerous non-myofibril cell populations such as satellite cells, stromal cells, immune cells, and endothelial cells. How exercising muscle communicates with these cells to initiate the remodeling process remains poorly understood. We performed comparative metabolomics analysis on exercised skeletal muscle to identify potential signaling molecules that preferentially increase in the local extracellular environment of muscle, following exercise. Using this approach, we identified succinate, a tricarboxylic acid (TCA) cycle intermediate, as a potential exercise signaling molecule in both mice and humans. Succinate is a ligand for a g-protein coupled receptor, succinate receptor 1 (SUCNR1). Using a hybrid resistance-endurance training model, we implicate succinate-SUCNR1 signaling in a host of remodeling processes such as fast-twitch myosin protein expression, innervation, and extracellular matrix remodeling. These changes manifest as a SUCNR1-dependent increase in strength. Additionally, we show that SUCNR1 is exclusively localized to non-myofibrillar cells within skeletal muscle. Moreover, we identify a novel pH-gated transport mechanism for succinate where monocarboxylate transporter 1 (MCT1) is repurposed to allow for succinate secretion from skeletal muscle, during exercise. We also show this transport mechanism to be relevant in brown adipose tissue (BAT) where succinate sequestration by BAT has been linked to many favorable outcomes including reduced systemic tissue inflammation and increased energy expenditure. Overall, we identify a novel role for succinate as an exercise-responsive metabolite, implicate succinate-SUCNR1 signaling in many muscle remodeling processes, and characterize a new transport mechanism for succinate
Expanding the repertoire of methodologies for CRISPR-based genetic manipulation and for hit compound discovery in Mycobacterium abscessus
No full textThe recent years have seen a rise in infections caused by nontuberculous mycobacteria (NTM), especially by the highly drug-resistant Mycobacterium abscessus. Its large arsenal of intrinsic resistance mechanisms renders most antibiotics ineffective and current treatment regimens are suboptimal in curing M. abscessus infections despite prolonged use of antibiotic cocktails. Some success has been made to develop alternate drug combinations based on repurposing of approved antibiotics, but progress to advance new therapeutic options is hampered by conventional drug screening failing to yield sufficient hits for further development or identifying hits against the same targets such as MmpL3. Additionally, while transposon-sequencing (Tn-seq) has been the mainstay of genome-wide functionality studies, tools for genetic manipulation of M. abscessus have been tedious and inefficient in generating the desired engineered strains due to low rates of homologous recombination and high rates of background spontaneous antibiotic resistance.
In Chapters 2 and 3, we demonstrate analogous two-plasmid workflows that incorporate use of a fluorescent mCherry reporter to identify desired strains easily. We apply the recently developed mycobacterial CRISPR interference (CRISPRi) in Chapter 2 to generate CRISPRi hypomorph strains rapidly in M. abscessus, and subsequently validate through targeted gene silencing essentiality calls made by FiTnEss and HMM, two complementary analytical methods for Tn-seq datasets. In Chapter 3, we convert this CRISPRi platform into one that can perform CRISPR/Cas9-mediated genetic disruptions and establish an analogous two-plasmid workflow that can generate simple targeted single gene disruptions, and more complex multiple gene disruptions with 102-104 higher efficiency than reported for existing recombination-based methods. These complementary tools will greatly expedite targeted genetic manipulation to expand our understanding of the biology, pathogenesis, and drug resistance mechanisms of M. abscessus.
To address the limitations of whole-cell phenotypic screening, we first establish a multiplexed, target-based phenotypic screening platform in Chapter 2, using CRISPRi for the first time to evaluate chemical-genetic interactions in high throughput for the purposes of hit discovery. With our pilot screen, we show our method can predict chemical-genetic interactions with reasonable accuracy (65%) and can identify potential weakly active hit compounds while giving insight into their potential mechanisms of action. Using InhA inhibitors as an example, we hypothesize that InhA may represent a relatively overlooked target for M. abscessus, and that even intrinsic resistance to a well-known anti-tuberculosis drug, isoniazid, is complex and multi-factorial. We then show in Chapter 4 that screening a biased library of anti-tuberculosis bioactives can lead to the identification of new hit candidates with underexplored mechanisms of action, as seen with the discovery of novel compounds predicted to inhibit FadD32 and IlvB1 in M. abscessus. We thus propose two alternate approaches for improved hit discovery. By expanding the target genes included in the hypomorph screening pool, and by expanding curated libraries of anti-tuberculosis bioactives to include diverse mechanisms of action, we can potentially enrich screening hits for novel compounds representing a diverse range of mechanisms
Scaling and Renormalization in Statistical Learning
No full textThis thesis develops a theoretical framework for understanding the scaling properties of information processing systems in the regime of large data, large model size, and large computational resources. The goal is to develop an understanding of the impressive performance that deep neural networks have exhibited.
The first part of this thesis examines models linear in their parameters but nonlinear in their inputs. This includes linear regression, kernel regression, and random feature models. Utilizing random matrix theory and free probability, I provide precise characterizations of their training dynamics, generalization capabilities, and out-of-distribution performance, alongside a detailed analysis of sources of variance. A variety of scaling laws observed in state-of-the-art large language and vision models are already present in this simple setting.
The second part of this thesis focuses on representation learning. Leveraging insights from models linear in inputs but nonlinear in parameters, I present a theory of early-stage representation learning where a network with small weight initialization can learn features without altering the loss. This phenomenon, termed silent alignment, is empirically validated across various architectures and datasets. The idea of starting at small initialization leads naturally to the "maximal update parameterization", μP, that allows for feature learning at infinite width. I present empirical studies showing that practical networks can approach their theoretical infinite-width feature learning limits. Finally, I consider down-scaling the output of a neural network by a fixed constant. When this constant is small, the network behaves as a linear model in parameters; when large, it induces silent alignment. I present theoretical and empirical results of the influence of this hyperparameter on feature learning, performance, and dynamics
Places Apart: Buddhist Reclusion in Medieval Japan
No full textThis dissertation evaluates the impact of reclusive monks (tonseisō) on medieval Japanese Buddhism. It combines archival and archaeological evidence to analyze the relationship between recluses and local society. I argue that reclusion was a social activity that involved interaction with people of all walks of life. To support this argument, I focus on places of reclusion – called bessho, or “separate places” – as sites for both reclusive and social practice. I begin with a historiographical approach to bessho. Rather than a type of temple institution, bessho evoked rhetorical distance. As places distant from monasteries and cities, bessho afforded both isolation and opportunities for recluses to interact with people of various social backgrounds. Turning to three case studies, I first show how materials from votive burial deposits from Mount Kurama illustrate a vibrant community of local infantrymen who engaged in ritual burial alongside reclusive monks. I then turn to the example of the reclusive monk Sainen and his hermitage, Daihizanji, from which he ministered to and incorporated the interests of agricultural families in Hanase bessho. The final chapter examines the Ōhara bessho, a center for reclusive monks, woodsmen, and charcoal kiln workers that became a cradle for the interaction between Buddhist practice and economic production. These examples demonstrate how places of reclusion saw a negotiation of Buddhism with medieval social life. The conclusion discusses how these case studies compel a reinterpretation of the ways in which Buddhism became a religion for the medieval Japanese populace and raises some directions for future research
Homiletic Lives of Irish Saints
No full textThe Latin accounts of the lives and works of Irish saints have been a mainstay of historical and literary scholarship on medieval Ireland for centuries. Their vernacular counterparts are, with few exceptions, far less studied.
This dissertation examines a particular subset of these vernacular Lives: namely, those given the form of a homily, and which focus on the Life of an Irish saint. These Lives, twelve in number, were designed for annual recitation on the feast-day of their subject. They begin with a reading of a pericope from the Bible, along with an allegorical interpretation of that passage; they then continue with a narration of the saint’s life and deeds, and conclude with a brief peroration expressing the hope that the speaker and audience should reach Heaven and dwell there for all time.
It has been argued that these Lives once formed part of a homiliary compiled at Armagh in the late eleventh century. This dissertation shows that this hypothesis is untenable. The close (often verbatim) similarities between these twelve Lives reflect, rather, the church of Armagh’s potent influence in the northern half of Ireland, as authors from Mayo to Meath used and adapted a template pioneered many years earlier in the hagiography of Patrick, Armagh’s patron saint.
To support this claim, I investigate the source material for the homilies, much of which, as I show, has roots in Irish exegesis composed in the early Middle Ages; many of the homilies, moreover, were reliant on works produced at Armagh in later centuries. I also demonstrate, for the first time, the relationships between the various manuscripts that contain these Lives at the level both of individual texts and of the collections in which they usually circulated. Finally, I provide fresh transcriptions of five Lives, all either previously unedited, or edited only from defective manuscripts. Taken together, the following chapters offer an overview of the entire lifespan of the genre, from the origins of the homilies in seventh-century exegesis down to the final stages of their reworking in the early eighteenth century
Nanoscale spin-polarized imaging of magnetic Weyl semimetal CeBi
No full textWeyl fermions are massless excitations with definite chirality. They manifest around band-touching points in the bulk band structure when inversion or time-reversal symmetry is broken. Among Weyl semimetals, those breaking time-reversal symmetry are of special interest due to their potential for the controlled manipulation of Weyl points through magnetic structure engineering. Here, we present a systematic investigation of the Weyl state in different magnetic phases of CeBi and report the discovery of a tunable magnetic Weyl semimetal state within nanoscale fully-polarized domains in the antiferromagnetic (++−−) CeBi.
In the ferrimagnetic (+++−) phase, quasiparticle interference (QPI) measurements show a 100 meV band splitting in the Bi 6p band, which supports the existence of Weyl points in this phase. Similarly, in the fully-polarized (++++) state, we observed at least a 200 meV band splitting in the Bi 6p band along the kx axis, supporting the existence of Weyl nodes in fully-polarized CeBi.
We demonstrate the creation of Weyl nodes by inducing local ferromagnetic (FM) domains in antiferromagnetic CeBi. These FM domains consist of co-aligned Ce moments and can be generated through in-plane magnetic field training or by employing a scanning tunneling microscopy (STM) tip to induce local strain. We image the formation of Weyl fermions around these FM domains by measuring QPI patterns. Our results not only demonstrate CeBi as an excellent magnetic Weyl semimetal for investigating intrinsic Weyl physics but also show the possibility of controllably writing the Weyl phase at the nanoscale.
Finally, we present a novel way to accelerate QPI measurements by acquiring sparsely measured dI/dV maps and full-grid topography maps simultaneously. The acquired full-grid topography maps facilitate the computation of drifting phase maps, essential for correcting lattice distortions within reconstructed dI/dV maps. Our results demonstrate the effectiveness of sparse sampling, thereby broadening the scope of QPI applications