116964 research outputs found
Sort by
Data-driven strategies for effectively using scenarios in single and multistage decision problems
Get PDFWhen the true model dynamics are unclear, scenarios are widely employed to model uncertainty in decision-making problems. Different issues arise when using scenarios in static and dynamic problems. This research looks at two challenging problems: scenario selection and dealing with ambiguity in observed scenarios. Consider a static situation in which a large number of scenarios are available and employing all of them is computationally challenging. What criteria should the decision-maker employ when selecting scenarios that provide the most information about the worst-case risk? How should a decision-maker select a small yet effective sample of scenarios from a large collection that is reflective of worst-case risk? This forms the core theme of the first study subject. The second topic takes the idea of coping with uncertainty in single-period settings and extends it to multiple periods. It’s unsurprising that using unrealistic scenarios can lead to incorrect models that fail to represent the underlying dynamics. Making sub-optimal decisions based on erroneous model dynamics can also be expensive. The major issue of the second line of research is dealing with uncertainty in the underlying model dynamics in multi-period decision-making in order to derive robust decisions that are optimal against moderate perturbations from the observed uncertainty. The research topics not only add to the body of knowledge in multiple areas, including scenario selection and distributionally robust multistage optimization, but also provide risk managers, policymakers, and decision makers with actionable insights. Stress tests based on extreme-yet-plausible scenarios have become a preferred approach of assessing risk for large financial institutions since the global Financial Crisis of 2008, but the majority of scenario selection has been subjective and unstructured. The first essay presents a principled framework for selecting scenarios by minimizing the risk of a financial institution’s profit and loss (P/L) distribution. A link between experimental design and stress testing is formed by adding a risk-based design target to the design of experiments framework. Three separate design criteria are presented. A parametric design, where the P/L depends linearly on risk factors; 2) a distributionally robust design, where the P/L depends linearly on risk factors whose distribution is assumed to lie within an uncertainty set based on the Wasserstein metric; 3) a non-linear design, when the P/L is a non-linear function of the risk factors. To demonstrate the value of the proposed framework, CVaR accuracy is evaluated using stress scenarios selected under the new framework and compared to CVaR accuracy using stress scenarios issued by the Commodity Futures Trading Commission in 2019 to stress test central counterparties. The second essay examines a multistage distributionally robust optimization framework that protects against a class of stochastic processes that are similar to the empirical stochastic process generated from the sample paths. Nested distance developed by Pflug and Pichler (2012) is used to determine the proximity of information dynamics. Because of the non-convexity of the distributional uncertainty set, the problem with nested distance is difficult to solve. A strong duality reformulation is used to generate computationally tractable dynamic programming reformulations under a variety of situations. The reformulations’ efficacy is demonstrated using the dynamic portfolio selection problem.Information, Risk, and Operations Management (IROM
Mapping the Black manosphere
Get PDFThis thesis introduces and delves into the concept of the "brown pill," a nuanced ideology akin to the red pill philosophy but with a distinct focus on men of color, particularly Black men, challenging conventional narratives within the manosphere. By blending elements of red pill ideology with rhetoric from Black nationalist movements, the brown pill complicates assumptions of privilege and gender. Through Critical Technological Discourse Analysis, this thesis examines how the brown pill intersects with race, gender, and digitality within the Black Manosphere. Case studies of influential figures such as Kevin Samuels, the "Fresh and Fit" podcast, and the passportbros movement shed light on how these communities negotiate Black masculinity online and address historical tropes of misogynoir and heteropatriarchy. This research fills a gap in literature by exploring the intersectionality of gender and race at this level and underscores the importance of understanding the evolution of disgruntled masculine performances by Black men and the remediation of far-right misogynoir within digital spaces.Radio-Television-Fil
Quantum computers and Monte Carlo estimation
Get PDFQuantum algorithms have a fascinating relationship with randomness. On the one hand, the dynamics of quantum circuits and quantum physics defy description with classical probability. This is, in some sense, because quantum mechanics requires the use of 'negative probabilities' to describe what is going on. On the other hand, it is these very same negative probabilities that lend quantum computers their power. While classical Monte Carlo strategies fail to efficiently simulate arbitrary quantum computations, quantum computers can be used to speed up the performance of Monte Carlo algorithms. In this dissertation we explore the connections between Monte Carlo estimation and quantum computation. We present several quantum algorithms for Monte Carlo estimation tasks in physics, such as estimating physical quantities and measuring observables. We also present classical algorithms that simulate quantum computers using Monte Carlo strategies that succeed in avoiding a sign problem for an interesting family of circuits. We begin by deepening our understanding of the seminal result by Brassard, Høyer, Mosca, and Tapp [BHMT00]: while classical Monte Carlo estimation requires O(ε⁻²) samples to estimate a quantity to accuracy ε, quantum computers merely require O(ε⁻¹). While this algorithm is extremely widely used, it is also very complicated. We give a new, simpler quantum algorithm for quantum estimation. We find that the problem reduces to estimating a parameter θ, given access to a coin with bias sin²(nθ). Next, we present a review of block encodings. Block encodings are a recently developed technique for designing quantum algorithms that permit us to use non-unitary matrices. We also review singular value transformation, a powerful technique for applying functions to the singular values of block encoded matrices. While [GSLW18] presents a comprehensive analysis of these techniques, we find that there is a need in the literature for a more accessible introduction to the subject. Several results in this dissertation rely on block encodings and singular value transformation. We proceed to use the power of block encodings to construct quantum algorithms for the estimation of some physical quantities. We give a subroutine that, given access to a block encoding of O can estimate the expectation <O>. This subroutine harnesses the quadratic speedup for Monte Carlo estimation presented earlier. We then build algorithms for computing n-time correlation functions, the density of states of a Hamiltonian, and linear response functions. Another fundamental task in simulating physical systems is measuring in the eigenbasis of an observable O and extracting an eigenvalue. This is usually performed via a very complicated method: we pretend that O is a Hamiltonian, perform Hamiltonian simulation e [superscript iOt], and then use phase estimation to extract the eigenvalues of O. We find that block encodings and singular value transformation let us build a much simpler algorithm for this task. The idea is to directly construct block encodings of matrices encoding the bits of a binary expansion of the eigenvalues of O, and then the extract the estimate one bit at a time. We find that this method is about 20x faster than the one based on phase estimation. Finally, we turn to classical Monte Carlo strategies for simulating quantum circuits. A famous early result of quantum information was that classical computers can simulate Clifford circuits [Gottesman98]. We combine this result with quantum Monte Carlo to obtain a classical algorithm for simulating all quantum circuits. However, the further away the circuit is from the Clifford group, the worse the impact of the sign problem becomes. These algorithms have several interesting properties. They can simulate Clifford circuits in linear time, which is quadratically faster than the best known deterministic algorithm [AG04]. They can also simulate a large family of input states that cannot be made using Clifford circuits.Physic
Tracing rustbelt mythologies
Get PDFDetroit, a city marked by labor exploitation, racial segregation, systemic inequality, and criminally underfunded infrastructure, which, in the form of sprawling, cracked road systems, carves out neighborhoods and divides rich from poor. Though the murals were completed nearly 90 years ago, this dissonance between representation and reality is not simply the result of changes in the city, though there have been many. Rivera embraced contradiction, accepting corporate commissions while representing communist ideals and imagining a utopian future in the face of dystopian conditions. Much has been written about the Mexican muralist’s revolutionary work, including leftist critiques of his relationships with American elite circles and his complicated time in Detroit, where he organized with Mexican laborers while dining with the Fords. While this research is invaluable, what I am primarily interested in is the murals' enduring place in metro Detroit's cultural imagination and identity. They not only depict contradicting, complex ideals, but weave diverse temporal threads, asking that we see the future and past concurrently. Their relevance today speaks to metro Detroit’s long-fomenting contradictions: the institutional and grassroots desire to create a utopian, collective identity that transcends race, nation, class, and time, and the enduring afterlives of Detroit’s segregationist history that are often absent from institutions’ mythologized narratives but evident outside them. My hope is that in tracing the revolutionary desires and industrialist myths in the Detroit Industry Murals, we can create narratives about ourselves that hold these contradictions to light.Comparative Literatur
Intersecting perspectives on intersecting alliances : youth, caregiver, and therapist perspectives on the alliances in youth psychotherapy and the unique contribution of therapist meta-perceptions
Get PDFTherapeutic alliances have been generally understudied in child and adolescent literature, including therapist alliances with both youth clients and caregiver stakeholders, who are intimately involved in initiating and maintaining their children in treatment. Nevertheless, extant research suggests that maintaining these relationships are highly relevant to effective therapeutic processes. Research also suggests that factors of the therapist and the therapist’s ability to accurately detect the functioning of the therapeutic alliance additionally contribute to therapy effectiveness, according to adult psychotherapy literature. The current study applies a truth-and-bias model previously only used in adult models of psychotherapy to a usual care sample of youth clients and their caregivers in outpatient therapy using data from two existing randomized control trials of a measurement feedback system. Secondarily, it uses structural equation modeling techniques to explore comparative reports of the alliance from youth and caregiver clients and their therapists and relate discrepancies to changes in child internalizing and externalizing symptoms. Results suggest that, like in previous studies, therapists’ estimate the alliance to be lower than their child and parent clients estimates. In addition, therapist meta-perception of the parent-therapist alliance was predicted by therapist report of the alliance, but not parent report of the alliance. By contrast, therapist meta-perception of the child-therapist alliance was predicted by both the therapist report of the alliance and child-report of the alliance. Adequately fitting structural equation models were developed to describe youth and therapist reports of the alliance, but no adequately fitting model described the relationship between caregiver and therapist reports of the parent-therapist report alliance. In addition, no model connecting reporters of the alliances to internalizing or externalizing models demonstrated adequate fit statistics, and paths between latent alliance variables and symptom slopes were not significant.Educational Psycholog
Electrostatic interactions at the Arf1-BFA-GEF interface : measuring cyanocysteine electrochromic effects to characterize a protein-drug-protein interface
Get PDFProtein-protein interactions regulate many cellular processes, making them ideal drug candidates. Design of such drugs, however, is hindered by a lack of understanding of the factors that contribute to interaction specificity. Specific protein-protein complexes possess both structural and electrostatic complementarity, and while structural complementarity of protein complexes has been extensively investigated, fundamental understanding of the complicated networks of electrostatic interactions at these interfaces is lacking, thus hindering rational design of orthosterically binding small molecules. To better understand the electrostatic interactions at protein interfaces and how a small molecule could contribute to and fit within that environment, we used a model protein-drug-protein system, Arf1-BFA-ARNO4M, to investigate how the small molecule brefeldin A (BFA) perturbs the Arf1-ARNO4M interface. We also investigated the electrostatic impact of a variety of guanine nucleotide exchange factors (GEFs) on the Arf1-GEF interface, to learn how BFA sensitive GEFs, Gea1p and ARNO4M, differ from BFA insensitive GEFs, ARNOwt, ARNO1M, and ARNO3M. By using nitrile probe labeled Arf1 sites and measuring vibrational Stark effect as well as temperature dependent infrared shifts, we measured changes in electric field and hydrogen bonding at this interface upon GEF or BFA binding. At all five probe locations of Arf1, we found that the vibrational shifts resulting from BFA binding corroborates trends found in Poisson-Boltzmann calculations of surface potentials of Arf1-ARNO4M and Arf1-BFA-ARNO4M, where BFA contributes negative electrostatic potential to the protein interface. The data also corroborate previous hypotheses about the mechanism of interfacial binding and confirm that alternating patches of hydrophobic and polar interactions lead to BFA binding specificity. Comparing BFA sensitive and insensitive GEFs showed many similarities between Gea1p, ARNOwt, ARNO1M, ARNO3M, and ARNO4M, suggesting an ideal structural and electrostatic environment for Arf1 binding and likely BFA specificity. These findings demonstrate the impact of BFA on this protein-protein interface as well as the level of specificity for a particular protein-protein interfacial environment, which have implications for designing other interfacial drug candidates.Biochemistr
Efficient derandomization and simulation of BPP under assumptions
Get PDFRandomized algorithms have proven extremely useful in both providing efficient solutions to computational problems, and extending our understanding of the limits of computation. But, in practice they have their downsides. First, the output of randomized algorithms is inherently uncertain. Second, perfectly uniform randomness may be hard to come by in nature. Classic results provide solutions to these issues. Under assumptions, we know that any polynomial time randomized algorithm can be derandomized into a deterministic algorithm running in time a much larger polynomial. Additionally, any polynomial time randomized algorithm can be simulated using weak, imperfect randomness, again, in time a larger polynomial. Large polynomial time, such as n¹⁰⁰ is undesirable in practice. This thesis demonstrates that efficient derandomization and simulation is plausible. First, under stronger assumptions, we show that any randomized algorithm running in time t has a deterministic counterpart running in time t². Second, we show that under a natural assumption about the structure of the imperfect randomness, we can use it to simulate randomized algorithms with virtually no slowdown. En route to these results, we demonstrate novel methods to bypass well known barriers in complexity that stand ubiquitously in the way of efficiency: the hybrid argument and the union bound. Towards our simulation result, we prove an intimately related result that shows that adversarial random walks using instructions that are neither uniform nor independent can still mix well.Computer Scienc
Obesogenic behavioral, metabolic, and neuronal consequences of adolescent social stress in male hamsters
Get PDFAdult male hamsters exposed to chronic social stress gain more weight, eat more food, and have more body fat than controls, a condition that mirrors stress effects in humans. It was unknown whether stress during adolescence caused similar results and, if so, what the longevity of those effects were. Male golden hamsters were exposed daily to aggressive adults from early to mid puberty – postnatal day 28 to 42. Body weight and food intake was tracked every two days. Over the two week period, Stressed subjects gained weight at a faster rate, which differentiated after only two days of stress, ate more food, had enhanced food efficiency, and ultimately weighed 10% more than controls, alongside a 15% increase in body fat. Stressed subjects also collected 50% more food in a food hoarding task on postnatal day 42, suggesting enhanced motivation for food at this stage. To investigate this, animals underwent the same stress protocol, and were also trained in a food conditioned place preference (CPP) task in the last five days. While all animals developed preference for the area previously associated with food, CPP was blunted by approximately 30% in stressed subjects. Additionally, to examine neural mechanisms likely involved in metabolic and other phenotypic traits associated with stressed hamsters, orexin-A reactive neuron fibers were quantified across the brain. There were no groupwise differences in orexin innervation in any area quantified. However, the correlation between orexin innervation in multiple hindbrain regions and various metabolic outcomes was inverted by stress. Finally, RNA Tag-seq analysis of the lateral, dorsomedial, and arcuate nucleus of the hypothalamus showed approximately 500 differentially expressed gene transcripts in each region in Stressed subjects. Differentially expressed genes were often related to metabolism, sex hormones, thermogenesis, RNA processing, synapse organization, and epigenetic modules. Thus, the present research emphasizes golden hamsters as a useful tool for future studies on stress, development, obesity, and other metabolic diseases, and proposed a number of potential mechanisms involved.Psycholog
Pedestrian-first corridors : studying the streetscapes of Barcelona's Superilla project
Get PDFIn this professional report, I will examine the global challenges associated with car-centric cities, exploring how various urban centers worldwide have responded to these issues and what led Barcelona to initiate significant change. The focus will be on the Superillas project, a transformative urban intervention in Barcelona, Spain. This report will discuss the physical and social elements that constitute the Superillas, analyzing how these pedestrian-first corridors have redefined urban spaces. The report will specifically investigate the social dynamics within three distinct neighborhoods where Superillas have been implemented. By focusing on these neighborhoods, the study aims to reveal the broader implications of the Superillas, highlighting both the tangible changes in the built environment and the intangible shifts in social cohesion and community well-being. My hope is that this report will inspire U.S. cities to reflect on the Barcelona Superilla model and its potential to enhance residents' quality of life, promote healthier environments, and foster more vibrant, connected communities.Community and Regional PlanningArchitectur
Personalized medicine : developing high-performance inverse algorithms for multi-species brain tumor growth models
Get PDFThis thesis is on inverse problems for the integration of partial differential equation (PDE) tumor growth models with medical images. Our focus is on brain tumors and in particular glioblastomas. Our main contribution is to propose, implement, analyze, and evaluate an inverse problem solution methodology for a patient-specific, multi-species brain tumor growth model using a single-snapshot multi-parametric magnetic resonance imaging (mpMRI) scan. To our knowledge this is the first work that uses a multi-species tumor growth model and accounts for the unknown healthy patient anatomy, the unknown tumor initial condition, and for unknown tumor growth model parameters. From a clinical perspective the proposed work aims to establish a richer biophysical model—compared to the state-of-the-art, that can be then be used for downstream clinical tasks. PDE models are used to describe the complex process of tumor growth, with the ultimate goal of enhancing our understanding the dynamics of disease progression. However, calibrating such PDE models from imaging data can be challenging. In our case the most significant challenges are the availability of a single time snapshot mpMRI, the structural ill-posedness of the underlying inverse problem, and the high computational costs. The ill-posedness is exacerbated by the fact that, in addition to unknown tumor growth parameters, both tumor initial conditions and the normal patient anatomy are unknown. The high computational costs are due to the fact that the inverse problem requires the solution of a 4D (space-time) coupled problem comprising (1) the forward multi-physics on linear reaction-diffusion-advection PDEs with variable coefficients in moving domains; (2) the adjoint problem, which is a linearization of the forward problem but solved backward in time; and (3) the gradient problem that couples the forward and adjoint in a 4D manner. In this thesis we propose an inverse problem formulation and numerical algorithms that address these challenges. We first analyze the ill-posedness of the problem concerning the model coefficients. We propose an algebraic regularization technique that samples the Hessian and reconstructs a subspace of the most informative directions. The computational costs are treated by using state-of-the-art solvers developed in the group in the past for the forward solver and the tumor initial condition inversion, and by incorporating finite difference based sensitivities for the multi-species PDE coefficients. The inverse problem with mass effect (mass-induced deformation due to tumor growth) incurs a high computational cost due to the linear elasticity equations involved. To avoid excessive memory costs, we opted to compute derivatives without adjoints. Instead, we use the sensitivity equations. Our inversion scheme achieves a tenfold increase in speed compared to gradient-free optimizers. This scheme is evaluated using synthetically generated data as well as large, publicly available clinical datasets. The evaluation includes comparisons with single-species reconstructions, mismatch error analysis, and an assessment of the overall algorithmic robustness. Our model demonstrates an average improvement of 7% in whole tumor reconstruction over the widely used single-species reaction diffusion model. Additionally, our approach considers sub-tumor species and their interactions, providing enhanced information regarding brain vascularization.Computational Science, Engineering, and Mathematic