Efficient Greenfield Mineral Exploration

Hall, Tyler Stanford University ProQuest Dissertations & These 2023 해외박사(DDOD)

소속기관이 구독 중이 아닌 경우 오후 4시부터 익일 오전 9시까지 원문보기가 가능합니다.

A growing population will require more metal in order to sustainably build a high quality of life. However, there have been fewer discoveries of mineral deposits due to constrained exploration budgets and confounding geology. There are few practical tools for decision-support in greenfield mineral exploration. This dissertation focuses on developing and testing methods for increasing the efficiency of mineral exploration, with a focus on optimal planning of boreholes. The global context of mineral supply and an overview for mineral exploration is provided in Chapter 1. The necessary elements for modelling decision-making problems are defined and formulations are explained in the context of mineral exploration. Reward functions such as those based on hypothesis falsification or efficacy of information are demonstrated on illustrative cases. High dimensional optimization problems, such as those in the subsurface, are oftentimes intractable. To address this issue, approximations are introduced and demonstrated. Chapter 3 presents a decision-making problem on a real case in Western Australia, formulated as a partially-observable Markov decision process (POMDP) and solved using Monte Carlo tree search (MCTS) with a belief-based reward function. Belief-based rewards are demonstrated to be more performant than the state-of-the-art solvers, and a sensitivity analysis of performance to the belief-based reward is performed. Furthermore, a sensitivity analysis of the optimal plan to a decision-maker's input preference demonstrates how the optimal first decision changes based on a volume threshold. Chapter 4 formulates greenfield exploration drilling using sequential value-of-information, and shows that the selection of a first borehole changes if the decision-maker directly encodes the option of continuing to a second borehole. Discussion of future research directions conclude the dissertation.

Advances in Data-Driven Financial Econometrics and Item Response Theory: Theory and Applications

Liu, Chenru Stanford University ProQuest Dissertations & These 2021 해외박사(DDOD)

소속기관이 구독 중이 아닌 경우 오후 4시부터 익일 오전 9시까지 원문보기가 가능합니다.

This thesis consists of results and methods in four research projects, in chronological order, leading to advances in financial econometrics and item response theory. The first two projects were started when I was a master's student at Peking University's HSBC Business School. They were developed much further at Stanford after I took machine learning and approximate dynamic programming courses, leading to Chapter 2 and 3. The next two projects began when I worked on the analysis of ant foraging data with Professor Gordon and Lai, branched out into educational testing and mental health analytics when I worked with Professor Lai and Lu on these applications of item response theory, and culminated in Chapter 4.Each of the subsequent chapters is self-contained, with its own introduction, background and literature review, together with discussion and concluding remarks, thus obviating the need for a Conclusion chapter.

Improving Cosmological Predictions: Baryonic Effects and Analytical Techniques in the Effective Field Theory of Large Scale Structure

De Braganca, Diogo Pinto Leite Stanford University ProQuest Dissertations & These 2023 해외박사(DDOD)

소속기관이 구독 중이 아닌 경우 오후 4시부터 익일 오전 9시까지 원문보기가 가능합니다.

The theory of cosmic inflation stands as one of the most captivating hypotheses in the realm of early universe physics, captivating the imaginations of physicists since its development in the 1980s. Originally conceived to elucidate the conspicuous absence of observed magnetic monopoles, this theory unexpectedly offered an elegant and straightforward explanation for the universe's flatness and the striking uniformity of the Cosmic Microwave Background (CMB) temperature in all directions.Inflation, at its core, postulates a phase of exponential expansion in the infancy of our universe, driven by one or more quantum fields. The quantum fluctuations within these fields manifest as the primordial seeds for the formation of cosmic structures. Remarkably, inflationary models make predictions regarding the statistics of these primordial density fluctuations, aligning harmoniously with the observations of the CMB.However, while the CMB provides invaluable insights into the early universe, it does not unveil the precise physics governing the process of inflation. This unexplored terrain can be parametrized by what we term "primordial non-Gaussianities" - a measure of the deviation from perfect Gaussian distribution in the primordial fluctuations. Primordial non-Gaussianities represent a ubiquitous prediction within various inflationary models, awaiting further scrutiny.The large-scale structure (LSS) of the universe, specifically the spatial distribution of galaxies ad dark matter, constitutes a crucial avenue for constraining non-Gaussianities. Unlike the twodimensional CMB, LSS furnishes a three-dimensional tapestry rich in information, encompassing an abundance of modes that hold an enormous amount of information. Developing a robust theoretical framework to analyze the LSS derived from galaxy surveys assumes paramount importance in our quest to extract the physics of inflation.The framework that has emerged as a result of diligent efforts by numerous physicists is aptly known as the Effective Field Theory of Large Scale Structure (EFTofLSS). This pioneering construct empowers us to probe the quantum fields that drove cosmic inflation from the universe's largest scales, surpassing the capabilities of any terrestrial particle collider.This thesis embarks on an intellectual odyssey dedicated to the development of the EFTofLSS, encompassing both theoretical and computational aspects. By bridging the gap between theory and observation, this work aims to unlock the secrets of the early universe and shed light on the fundamental nature of our cosmos.Chapter 1 serves as the gateway to this voyage, providing a general introduction to cosmology. Chapter 2 provides a review of cosmological perturbation theory and the EFTofLSS. Subsequently, Chapter 3 takes center stage, delving into the incorporation of normal, visible matter into the EFTofLSS - an endeavor that expands upon the previous formulation developed solely for a single dark matter component. This significant advancement has been made possible through fruitful collaborations with esteemed researchers, including Matt Lewandowski, Leonardo Senatore, Raphael Sgier, and David SekeraAs we traverse deeper into the realm of the EFTofLSS, Chapter 4 unveils an efficient analytical methodology for calculating loop corrections, enhancing the possibility of using this framework to perform cosmological inference. This new approach has emerged through fruitful collaborations with Henry Zheng, Babis Anastasiou, and Leonardo Senatore.

Data-Driven Statistical Sharding for Industry-Scale Neural Recommendation

Sethi, Geet Stanford University ProQuest Dissertations & These 2023 해외박사(DDOD)

소속기관이 구독 중이 아닌 경우 오후 4시부터 익일 오전 9시까지 원문보기가 가능합니다.

Deep learning based recommendation models (DLRMs) form the backbone of many internet-scale services such as web search, social media, and video streaming. Primarily composed of massive embedding tables, potentially terabytes in size, these models require immense system resources to train and the solving of the sharding problem. The sharding problem is the task of partitioning and placing the embedding table parameters throughout the target system memory topology such that training throughput is maximized.This dissertation: (1) Characterizes and derives statistics from DLRM training data which can be used to accurately and granularly predict the memory demands of individual embedding table rows; (2) Presents RecShard, a mixed-integer linear program based approach which uses these statistics to solve the sharding problem for capacity constrained single-node systems, where parameters must be placed across high-performance GPU HBM and much slower CPU DRAM; reducing accesses to the latter by orders of magnitude; and (3) Presents FlexShard, a precise row-level sharding algorithm which focuses on sharding emerging sequence-based DLRMs across multi-node GPU training clusters; leveraging these statistics to significantly reduce inter-node communication demand, the bottleneck of scale-out DLRM training.The size of industry-scale DLRMs requires sharding to be performed; however the skewed power-law nature of DLRM training data causes imprecise partitioning and placement decisions to result in imbalanced load across the system memory topology. The contributions of this dissertation provide a foundation upon which one can reason about the access patterns to fine-grained regions of DLRM memory; as well as two novel sharding techniques built upon this foundation. These techniques demonstrate significant improvements over the prior state-of-the-art on real-world production data and system deployments.

Death is a Feast: The Ideology of the Homeric Banquet

Tewksbury, Ian Arthur Stanford University ProQuest Dissertations & These 2023 해외박사(DDOD)

소속기관이 구독 중이 아닌 경우 오후 4시부터 익일 오전 9시까지 원문보기가 가능합니다.

In large part, feasting in the Iliad and the Odyssey has been critically overlooked by studies on the political ideology of the Homeric poems. This dissertation studies the meaning of the feast, δαίς, in these poems and contends that the feast in Homeric poetry is what the anthropologist Bryan Hayden has called "a critical element, almost entirely overlooked in the past." Feasting, as the anthropologists Bryan Hayden and Michael Dietler have shown, forms critical reciprocal obligations which encompass "the political, economic, ritual, and social universe." Likewise, I argue that feasting reveals the reciprocal obligations at the heart of the political, economic, and ritual world of Homeric society. By providing an updated formulaic and linguistic analysis of the language of feasting in Homer, the "critical" social relations of the Homeric community are revealed. Part I one of this dissertation provides an exhaustive formulaic analysis of two contrastive verbs used to describe feasting: (1) δατέομαι, "to divide up, share, " and (2) δαίνυμι, "to feast." Part II analyzes the formulaic meaning of the δαίς, "feast, " as a particular conceptual form of division. I show that the δαίς, derived from the Indo-European root, *deH2-, "to divide, " is not a "meal" that embodies egalitarian participation, but is instead a gift that instantiates an ideal form of hierarchy (social division). This study resolves contradictory definitions in the Homeric lexicography of feasting by showing how they rely on incomplete and partial conceptualization of the Homeric idea of division. Finally, this study concludes by re-contextualizing the Homeric feast within the anthropological study of feasting. It contends that the reciprocal obligations of the feast reflect a singular vision of ideal power and community: single patron rule. This allows us a unique perspective on the question of the political ideology of Homeric poetry. Instead of the struggle between aristocratic and plebiscitary democratic ideology, it is suggested that the Homeric feast reflects an ideological struggle between royal tributary rule (Odysseus' patron feast) and the rise of oligarchy (the Suitors' symposium) in the 8th to 6th century BCE.

Cosmic Shear, Or Atmosphere? Modeling the Atmosphere for Dark Energy Measurements with the Vera C. Rubin Observatory

Hebert, Claire-Alice Stanford University ProQuest Dissertations & These 2023 해외박사(DDOD)

소속기관이 구독 중이 아닌 경우 오후 4시부터 익일 오전 9시까지 원문보기가 가능합니다.

The Vera C. Rubin Observatory Legacy Survey of Space and Time (LSST) will image tens of billions of galaxies over ten years–a 100 times increase over previous surveys–enabling the most precise measurements of dark energy properties to date. Realizing the full potential of this powerful dataset requires an unprecedented understanding of observational effects that, if uncorrected, can bias the science results. One such source is the blurring of images due to optical turbulence in the atmosphere, which dominates the point-spread function (PSF) for ground-based instruments. This atmospheric effect imprints spatially correlated noise on scales (and with amplitudes) similar to the cosmological signal we will study: the spatial correlation of positions and shapes of galaxies on the plane of the sky due to gravitational lensing of light by the dark matter in the Universe. High-fidelity simulated astronomical images are an important tool in developing and measuring the performance of image-processing algorithms that will be needed to accurately and precisely account for sources of correlated noise such as the atmospheric PSF.This thesis presents a new simulation tool, psf-weather-station, that allows us to study and model the dependence of correlations in the atmospheric PSF on weather conditions at any observatory by leveraging data from weather forecasting models. We use this tool to simulate and study the correlations in the size and shape of the atmospheric PSF predicted for the Vera C. Rubin Observatory in Chile. We make quantitative predictions for two-point correlation functions (2PCF) that are used in analyses of cosmic shear. We observe a strong anisotropy in the two-dimensional 2PCF, which is expected based on observations in real images, and study the dependence of the orientation of the anisotropy on dominant wind directions at the observatory site.We also explore the temporal behaviour of the atmospheric PSF using high-resolution speckle images of stars, recorded near Rubin Observatory, and simulate similar observations using psfweather-station. We find differences in the time scales of correlation in the PSF between data and simulations, but observe a consistent dependence of PSF shape with image motion on short time scales.

Moral Cognition in Law and Policy

Petrov, Philip Stanford University ProQuest Dissertations & These 2023 해외박사(DDOD)

소속기관이 구독 중이 아닌 경우 오후 4시부터 익일 오전 9시까지 원문보기가 가능합니다.

The essays that comprise this dissertation study how the moral cognition of legal decisionmakers (judges, jurors, legislators, litigators, prosecutors, etc.) affects their behavior in the domains of criminal and tort law. The dissertation is inter-disciplinary and draws on research in law, politics, psychology, and philosophy. It proceeds by assembling existing experimental results and theoretical postulates into an original set of testable verbal hypotheses or theories. Chapter One considers criminal sentencing. It argues that, due to regularities in the moral psychology of blame, legal decision-makers are often motivated to use proportionality to choose appropriate prison sentence lengths, but they typically encounter difficulty when they proceed to calculate a proportionate sentence for a given crime. That proportionality is both cognitively intuitive and difficult to apply to prison sentences helps account for data points such as the following: why the original U.S. Sentencing Commission tried to, but did not, base the U.S. Sentencing Guidelines on a retributivist rationale; why sentencing decision-makers are likely to have politicalrhetorical flexibility in deciding whether to use the concept of proportionality; and why several federal judges have observed that sentencing decision-makers are susceptible to anchoring. Chapter Two, which turns to tort law, argues that dual-process theory in psychology partly explains tort’s justice-efficiency debate: whereas a set of relatively automatic mental representations preferentially supports the “corrective justice” view of tort, a different set of relatively controlled representations preferentially supports the “efficiency” view thereof. Understanding the justice-efficiency debate in dual-process terms partly explains, for instance, why the corrective justice view is historically older than is the efficiency view, why jurors and legal economists often clash in tort suits, and why the debate often takes the form of the question of whether tort doctrine’s reliance on moralized language is meaningful or perfunctory. Chapter Three, which continues the dissertation’s analysis of tort, argues that judges faced with the difficult task of drawing lines between legally cognizable and non-cognizable harm rely, reasonably, on the physical-emotional distinction to help them complete this task. Understanding tort’s physical-emotional distinction as a heuristic that helps judges solve a difficult problem undermines the argument, which some legal scholars have made, that judges who maintain the distinction thereby engage in mind-body dualist or otherwise unscientific thinking. The chapter’s analysis has at least three normative implications: users of tort’s physicalemotional distinction should clarify that they neither endorse dualism nor depreciate emotional harm; because judicial expertise may not extend to the task of drawing lines between legally cognizable and non-cognizable harm, judicial performance in this area may be more adequate than critics suggest; and, although it may not be possible to determine the optimal way of drawing lines between legally cognizable and non-cognizable emotional harm, moral-philosophical tools such as Rawlsian and Scanlonian contractualism may be able to identify partial or pro tantoconsiderations for choosing among different ways of doing so.

Measuring and Teaching Problem-Solving Practices in Digital Learning Environments

Wang, Karen Dan Stanford University ProQuest Dissertations & These 2023 해외박사(DDOD)

소속기관이 구독 중이 아닌 경우 오후 4시부터 익일 오전 9시까지 원문보기가 가능합니다.

Digital learning environments are becoming increasingly ubiquitous as a wide range of EdTech products and services enter classrooms and households across the globe. One salient attribute of these environments is their capacity to generate large amounts of data as students interact with the technology. These data logs can help construct a detailed picture of how students work on a task and provide valuable insights into their underlying competencies. At the same time, the sheer volume of interaction data poses challenges, such as how to extract meaningful behavioral patterns from the raw data and model them to assess specific constructs. This dissertation contributes to the efforts of educational researchers and practitioners in harnessing the data generated by digital technology to support teaching and learning, with an emphasis on using interactive tasks to assess and teach problem-solving practices.The Introduction chapter reviewed the historical development and various lines of inquiry in problem-solving research. The chapter also provided a summary of our lab's previous work on problem-solving practices and expert decision-making. The goal is to contextualize the present research on log data-based measurement of problem-solving within the broader problem-solving research, thereby establishing its relevance and contribution to the field.The first article, "Using Prolific as a Data Collection Tool for Educational Research," explored whether Prolific, an increasingly popular online crowdsourcing research platform, could be used to collect data on college students' problem-solving performance. Specifically, the study employed an interactive problem to compare the task engagement and performance of college students recruited from the Prolific platform with those of college students enrolled in an introductory physics course in a large public university. The Mystery Gift problem challenges students to determine the weight of an unknown gift using known weights and a marked seesaw in the PhET simulation. Results show that Prolific participants performed on par with college students from the physics class in obtaining the correct solutions. Furthermore, college students who submitted incorrect answers were more likely than Prolific participants to make rushed cursory attempts to solve the problem. These results suggest that Prolific is a valid data collection platform for studying how college students solve complex, interactive problems in science and engineering domains.The second article is "Applying Log Data Analytics to Measure Problem-Solving." This study investigated the potential of log data, which recorded students' interactions with the digital task environment while solving the Mystery Gift problem, as evidence of their problem-solving practices. Building upon the validity of Prolific as a data collection tool established in the first study, a national sample of 80 US college students majoring in STEM fields was recruited via Prolific to participate in a 30-min online study. The log data was processed to reveal the sequence of test trials conducted by individual participants when solving the problem and the pauses between test trials. Our analyses revealed that taking deliberate pauses (10 secs) during problem-solving was a significant predictor of participants' problem-solving success and an indicator of specific problem-solving practices. The results highlight the value of log data in offering unobtrusive observations of students' problem-solving processes and the power of learning analytics techniques in extracting semantically meaningful features associated with specific problem-solving practices.

Green Is the New Ivory: the Rise of the Sustainability Imperative in U.S. Higher Education

Gavrila, Simona Gabriela Stanford University ProQuest Dissertations & These 2023 해외박사(DDOD)

소속기관이 구독 중이 아닌 경우 오후 4시부터 익일 오전 9시까지 원문보기가 가능합니다.

As central social and political institutions, U.S. universities have an important public service calling that extends beyond their teaching and research mandate. Over the past two decades, increasing numbers of U.S. universities have reoriented their institutional priorities towards becoming more sustainable across the board, which includes "greening" their physical campuses as much as their curricula. This dissertation investigates the role of U.S. universities in the global sustainability movement, as well as the impact of the movement itself on the internal organization of universities. The findings suggest that the proliferation of sustainability-affirming practices and metrics between 2005-2020 has contributed to the emergence of a new and distinct imperative in U.S. higher education, which manifests itself through a variety of organizational pathways and is also impacted by the interplay between campuses and their neighboring communities (known as "town-gown" relations). The sustainability imperative has the potential to reaffirm the prominent role of higher education in American society and is likely to reshuffle higher education rankings as the next generation of students will increasingly seek out sustainability-conscious universities over those lagging in that respect. While this increases the risk of greenwashing on some campuses in the short term, the public valorization of sustainability is a positive development for higher education overall, and a win for environmentalism in the long term. In particular, as elite colleges and universities in the U.S. strive to remain competitive, they can no longer afford to be seen as "ivory towers" and instead must strive to become "green beacons" within their communities, setting examples of sustainability innovation and performance for other institutions, both within and beyond the higher education sector.

Knowledge Work Reimagined: Lyric Life Writing and the University

Salvidea, Gabriela Lila Stanford University ProQuest Dissertations & These 2023 해외박사(DDOD)

소속기관이 구독 중이 아닌 경우 오후 4시부터 익일 오전 9시까지 원문보기가 가능합니다.

Postwar academic culture has seen the rise of a literary consciousness and an autobiographical mode. One instantiation of this mode is a contemporary style of writing called “autotheory.” While some scholars interpret autotheory as a para-academic literary phenomenon, wishing to trace its lineage outside the university, I show that it is, in fact, rooted in the postwar university, and should be seen as a challenge to prevailing mechanistic epistemologies within the university. As a third way between strictly creative and strictly scholarly writing, autotheory works against prohibitions of self and self-assembly found in a host of humanistic discourses, including philosophy, theory, literary studies, and creative writing. However, I also show that autotheory is a direct outcome of philosophy, specifically the ordinary language philosophy of Ludwig Wittgenstein and Stanley Cavell, and that it plays out key elements of that philosophy. The self-assembly that is accomplished by autotheory is neither narcissistic nor “neoliberal.” It is an antidote to the radical self-displacements found in knowledge work. These cultures of radical self-displacement, engaged in what I call the search-for-method paradigm, undermine actual insight in favor of scientism and the administration of concepts. In contrast to this, autotheoretical works, and the tradition of lyric philosophy from which they stem, can be seen as recuperating personality, the experience of meaning, and a coherent political consciousness.