DSpace@RPI (Rensselaer Polytechnic Institute)
Not a member yet
6809 research outputs found
Sort by
Becoming irradiated: the epistemic politics of neglect along india’s nuclear fuel cycle
No full textDecember 2022School of Humanities, Arts, and Social SciencesBecoming Irradiated demonstrates that the dual epistemic conditions of self-reliance and nuclear safety neglect experiences of radioactive contamination across three varied irradiated facilities in India that comprise the nuclear fuel cycle under study. The Nuclear fuel cycle, as the study’s technoscientific apparatus, puts things, facilities, distinct entities, bodies and illnesses that are out of relations, back into relations and, capture the self-reliance conditions that shape India as nuclear country. The facilities include Kudankulam Nuclear Power Plant, Tummalapalle Uranium Mine and Mill and the Mayapuri Scrap Metal Market modelled on developmentalist, technological and neoliberal self-reliance. The dissertation provides a map of self-reliance, exploring its multiplicity, to demonstrate the psychopathology of colonialism embedded in India’s conceptual practice of self-reliance. Through a discourse analysis of IAEA regulatory episteme, the dissertation’s regulatory intervention questions the epistemic premise of today’s international nuclear regulation—nuclear safety achieved in the facility’s technologies, operations and management protects health and environment from irradiation. It analyzes the ontological enactment of the nuclear safety episteme in the facilities, and critiques it through embodied, subjugated epistemologies of becoming irradiated around it based on situational, para-sited, sensory and multispecies ethnographic analysis. The dissertation is theoretically framed by Cognitive Justice which provides a space for thought experiments on how different knowledge systems from Global North and Global South come to coexist in tense and dialogical relations. Under Cognitive Justice creating knowledge is not a task marked off to technoscientific expertise and hence, in this dissertation, I treat experts and epistemically oppressed peoples who experience irradiation as “epistemic bodies” whose epistemic process is embedded in ways they relate with varied technologies. By putting the embodied knowledges of irradiation and the epistemology of nuclear safety that shapes regulation in the facilities in conversation, this dissertation demonstrates how nuclear safety with its material arrangements emerge as an epistemology of neglect, making neglect less of a moral/behavioral matter and more of knowledge issue in Agnotology. The embodied knowledges of radiation illnesses, the dissertation demonstrates, are made and legitimated by victims of irradiation through accounts shared in popular media and in village meetings, forging radioactive kinships—relations that are forged due to ionizing effects of irradiation. Through epistemic bodies, nuclear neglect, radioactive kinship and nuclear differentiation, this dissertation lays the grounds for Critical Nuclear Studies (CNS). The attention to the production and sharing of subjugated knowledges and as they challenge the official discourses of the nuclear order are components of the field of Critical Nuclear Studies.
This dissertation pushes forward the frame of cosmopolitical and relational thinking by critiquing the move away from epistemology to ontology in STS and pivots STS, Study of Expertise, Cognitive Justice, Southern theory and Critical Nuclear Studies along the relational turn. The analyzes puts forth a demand for a regulatory apparatus namely Embodied Radiation Protection regime that attends to the epistemic agency of local people living around the facilities in knowing radioactive contamination. The embodied radiation protection regime relies on Ariviyal as pluralized embodied/specialized knowledge, alter-tracers as sensory apparatus, and is founded on the porous relations established with radioactivity on neglected epistemological grounds of nuclear operations and becoming irradiated. This dissertation frames sites of evasion as places where responsibility unevenly gets distributed onto the victims under an embodied radiation protection regime if the state is not held accountable in an increasingly capital centered regulatory apparatus.Ph
Towards a Proper Evaluation of Automated Conversational Systems
No full textEfficient evaluation of dialogue agents is a major problem in conversational AI, with current research still relying largely on human studies for method validation. Recently, there has been a trend toward the use of automatic self-play and bot-bot evaluation as an approximation for human ratings of conversational systems. Such methods promise to alleviate the time and financial costs associated with human evaluation, and current proposed methods show moderate to strong correlation with human judgements. In this study, we further investigate the fitness of end-to-end self-play and bot-bot interaction for dialogue system evaluation. Specifically, we perform a human study to confirm self-play evaluations of a recently proposed agent that implements a GPT-2 based response generator on the Persuasion For Good charity solicitation task. This agent leverages Progression Function (PF) models to predict the evolving acceptability of an ongoing dialogue and uses dialogue rollouts to proactively simulate how candidate responses may impact the future success of the conversation. The agent was evaluated in an automatic self-play setting, using automatic metrics to estimate sentiment and intent to donate in each simulated dialogue. This evaluation indicated that sentiment and intent to donate were higher (p < 0.05) across dialogues involving the progression-aware agents with rollouts, compared to a baseline agent with no rollout-based planning mechanism. To validate the use of self-play in this setting, we follow up by conducting a human evaluation of this same agent on a range of factors including convincingness, aggression, competence, confidence, friendliness, and task utility on the same Persuasion For Good solicitation task. Results show that human users agree with previously reported automatic self-play results with respect to agent sentiment, specifically showing improvement in friendliness and confidence in the experimental condition; however, we also discover that for the same agent, humans reported a lower desire to use it in the future compared to the baseline. We perform a qualitative sentiment analysis of participant feedback to explore possible reasons for this, and discuss implications for self-play and bot-bot interaction as a general framework for evaluating conversational systems
Beyond Labels: Empowering Human with Natural Language Explanations through a Novel Active-Learning Architecture
No full textData annotation is a costly task; thus, researchers have proposed low-scenario learning techniques like Active-Learning (AL) to support human annotators; Yet, existing AL works focus only on the label, but overlook the natural language explanation of a data point, despite that real-world humans (e.g., doctors) often need both the labels and the corresponding explanations at the same time. This work proposes a novel AL architecture to support and reduce human annotations of both labels and explanations in low-resource scenarios. Our AL architecture incorporates an explanation-generation model that can explicitly generate natural language explanations for the prediction model and for assisting humans' decision-making in real-world. For our AL framework, we design a data diversity-based AL data selection strategy that leverages the explanation annotations. The automated AL simulation evaluations demonstrate that our data selection strategy consistently outperforms traditional data diversity-based strategy; furthermore, human evaluation demonstrates that humans prefer our generated explanations to the SOTA explanation-generation system
Control of tollmien-schlichting waves on a natural laminar airfoil via dynamic surface modulation
No full textMay 2023School of EngineeringThe study presented here demonstrates experimental mitigation of Tollmien-Schlichting (TS)waves on Natural Laminar Flow (NLF) wind tunnel models. Experiments were first conducted
on an unswept NLF airfoil at a chord-based Reynolds number of 9:90 x 10^5 and then
on a 30 degree swept-back NLF airfoil at a chord-based Reynolds number of 8:45 x 10^5. Control
of TS waves was facilitated by dynamic surface modifications using piezoelectrically-driven
oscillating surface (PDOS) actuators. On each model, the actuators were located at three
streamwise locations on the airfoil suction side. For TS wave control, a disturbance was
introduced to the flow using the upstream actuator, which phase-locked the TS waves. The
downstream actuators were used to mitigate the induced TS waves by introducing anti-phase
disturbances with the proper amplitude. The disturbances included either single frequency
or multi-frequency input waveforms. The TS waves were mitigated in both open- and closedloop
control schemes for the unswept model, and in open-loop for the swept-back model. In
open-loop control, the results demonstrate that dynamic surface modification can be used
to mitigate the TS waves for varying frequency bandwidths even in the presence of a large
adverse pressure gradient. In addition, a closed-loop scheme, using an iterative learning
control algorithm, succeeded in reducing the induced disturbance amplitudes to 11% of their
original values. The experiments demonstrated the ability not only to mitigate the waves
but also to amplify them if transition to turbulence is desired. With actuators implemented
on the swept-back configuration, experiments indicated that this flow control method feasibly
suppressed the Tollmien-Schlichting waves in a similar fashion. In addition, for the
unswept model without actuation, a separation bubble was identified near the third PDOS
actuator. Preliminary results showed that using the upstream PDOS yielded mitigation of
the separation bubble.M
Practicing on the platform: an autoethnography of presence in the zoom-mediated ashtanga yoga 'mysore room'
No full textAugust 2023School of Humanities, Arts, and Social SciencesThe global Covid-19 pandemic has caused massive disruption to social practices of nearly every kind, impelling groups and institutions to migrate in-person activities onto digital videoconferencing platforms such as Zoom. A major concern within contemporary media studies is to understand how this now-paradigmatic form of communication is reconfiguring “communities of practice” as they move onto the platform and seek to recreate the felt, collective “presence” of embodied interaction. However, little attention has been paid to interactive physical practices, including athletics, dance, and various somatic therapies that rely on non-verbal forms of communication, including touch and breathwork. My "hybrid autoethnographic" study focuses on Ashtanga Vinyasa Yoga, a method of modern postural yoga practice that has, since the 1970s, expanded from its point of origin in Mysore, India to become one of the most prominent transnational practice systems. By undertaking extensive participant observation in the hybrid-reality "Mysore Room" and ethnographically tracking how the broader community has adapted to rolling cycles of respiratory crisis, I situate the phenomenon within a history of "transomatic mediations" that have carried the method as body-to-body knowledge across a range of media technologies.Ph
deep neural network quantifies individual cardiovascular disease risk
No full textAugust 2023School of EngineeringCardiovascular diseases (CVDs) are the leading cause of global mortality, emphasizing the urgent need for accurate, accessible, and interpretable risk assessment methodologies. This dissertation introduces a deep learning pipeline aimed at individualized CVD risk quantification. The proposed pipeline integrates risk factors derived from low-dose computed tomography (LDCT) with conventional tabular risk factors. It seeks to harness the rich, albeit noisy, data provided by LDCT images effectively. This pipeline is designed to contribute to the clinical field by offering CVD risk assessments that are accessible owing to their reliance on widely available LDCT imaging and readily available patient information, and that are also interpretable for medical professionals, thus facilitating timely interventions. The venture of applying deep learning to CVD risk assessment is relatively unexplored, and in the process of crafting this pipeline, three key questions emerged: 1) Can LDCT provide sufficient information for deep learning models to effectively quantify CVD risk? 2) How can risk factors be embedded in a high-dimensional space to maximize deep learning capabilities? 3) How can diverse risk factors be integrated in an interpretable manner? These questions mark critical barriers in the path of developing a viable deep learning pipeline for CVD risk assessment. The dissertation is structured around three specific aims, each corresponding to one of these questions. Aim 1 introduces Tri2D-Net, a deep learning model that efficiently extracts CVD-related features from 3D LDCT images using three orthogonal 2D views, demonstrating impressive performance in CVD screening and risk quantification. Aim 2 presents the Regression Metric Loss (RM-Loss), a novel loss function that guides a deep learning model to generate interpretable high-dimensional embeddings of risk factors, thereby bridging the gap between the low-dimensional space of risk factors and the high-dimensional space where deep learning thrives. Aim 3 brings forward MIX-CVD, a model based on the Mixture of Experts (MoE) approach. MIX-CVD adeptly analyses the complex relationships among risk factors, generating highly accurate predictions of CVD risk while maintaining interpretability through an adaptive assignment of explicit weights to each risk factor based on an individual's condition. Together, these components form a pipeline that successfully addresses the key questions, providing an effective, accessible, and interpretable tool for individualized CVD risk quantification. By offering accurate predictions that are easy to interpret and rely on widely accessible LDCT imaging, this work could potentially advance the current state of CVD risk assessment, underscoring the potential of deep learning in enhancing clinical decision-making and patient care.Ph
Music making as assimilation: the practice of group listening
No full textMay 2023School of Humanities, Arts, and Social SciencesGroup Listening is a collaborative musical performance practice that I organized and led from 2017 to 2021 with a group of musicians under the name Ensemble Consensus. Group Listening was the conceptual term I coined to be referenced by Ensemble Consensus members as a mutually understood, overarching framework when designing and discussing our own projects. Notably, Ensemble Consensus was a group dedicated solely to Group Listening projects. At its core, Group Listening is a creative methodology for investigating various relationship dynamics via iterative activities that necessitate collective authorship, participation, and improvisation. Specifically, every Group Listening project has resulted in Ensemble Consensus members co-writing text-based guidelines that outline the ways we should relate to one another and our creative tools. These guidelines determined how we rehearsed, rather than what we played. Thus, our various public performances and offerings ended up looking, sounding, and feeling like rehearsals. With each project, Ensemble Consensus aimed to imagine different ways of facilitating social interaction to deepen our creative capacity for collaborative music making. Understanding Group Listening necessitates understanding the social circumstances that initiated the performance practice. Group Listening and Ensemble Consensus came about as a reaction to my lived experiences with assimilation and rootlessness. My migratory background as a Seoul-born, Shanghai-raised artist currently based in New York informs my understanding of music as a social activity, where explicit and implicit etiquettes and rules of collaboration inform how people engage with one another socially and musically. Analyses of a portfolio of projects undertaken by Ensemble Consensus demonstrate how Group Listening treats music making as an assimilatory process that can manufacture norms and manipulate relationship dynamics in specific ways, parallel to the experience of cultural assimilation. In doing so, I advocate for the acceptance of rootlessness as a meaningful condition that encourages multiplicity and fluidity within notions of selfhood and identity. Accepting rootlessness at the personal and interpersonal level illuminates how it can be applied at the cultural level. In this project, I apply the connections between rootlessness, assimilation, and Group Listening to the contemporary cultural interpretations of Koreanness. In so doing, I situate the musical practice within the history of my place of birth and clarify the origins of my desire to legitimize rootlessness and assimilatory practices like Group Listening.Ph
Extensions of the discontinuous galerkin difference method
No full textDecember 2022School of EngineeringThis thesis explores extensions to the discontinuous Galerkin difference (DGD) method as a means of simulating hyperbolic conservation laws in a stable and efficient way. The first part of the thesis focuses on an entropy-stable discontinuous Galerkin difference (DGD) method for hyperbolic conservation laws on unstructured grids. The entropy-stable DGD method takes advantage of existing theory for entropy-stable (diagonal-norm) summation-by-parts (SBP) discretizations. In the case of entropy-stable discretizations, the entropy variables rather than the conservative variables must be interpolated to the SBP nodes. A fully-discrete entropy-stable scheme is obtained by adopting a relaxation Runge-Kutta version of the midpoint method. In addition, DGD matrix operators for the first derivative are shown to be dense-norm SBP operators. Numerical results are presented to verify the entropy-stability of the DGD discretization in the context of the Euler equations. Accuracy studies reveal that the DGD method is efficient; indeed, like tensor-product DGD schemes, the unstructured DGD method exhibits superconvergent solution error for periodic problems. An investigation of the DGD spectra shows that the spectral radius is relatively insensitive to discretization order. Furthermore, the DGD scheme is applied to a one-dimensional Riemann problem, and global conservation and convergence in the L1 norm are observed. The second part of the thesis presents a generalized DGD (GDGD) method, where the basis functions are not associated with the element centers. In addition, generalized DGD basis that are enriched with non-polynomial functions are investigated. Two practical issues that cause the GDGD stencils to produce ill-conditioned Vandermonde matrices are addressed. Numerical experiments verify the accuracy of the GDGD interpolation operators and demonstrate that the GDGD spatial discretization exhibits superconvergent solution error. Finally, the flexibility of the GDGD method is exploited for r-adaptation. This adaptation method uses an optimization approach in which the objective function is the 2-norm of the discontinuous Galerkin residual and the optimization variables are the basis center locations. A gradient-based optimization algorithm is adopted, and details on the derivation and computation of the gradient using adjoint method are discussed. Numerical experiments demonstrate that the GDGD r-adaptivity method is capable of reducing solution and/or functional error by aligning the GDGD basis distribution with solution features.Ph
Efficient finite difference schemes for wave equations : part 1 : incompressible linear elasticity part 2 : hierarchical high-order accurate schemes
No full textDecember 2022School of ScienceThis thesis is comprised of two major contributions. The first is the formulation of an efficient finite difference scheme for time-dependent incompressible linear elasticity on complex geometry.
The governing equations are solved in displacement-pressure form to second-order accuracy in
space and time. A fractional-step approach is taken so that the time update for the displacement is
performed separately from the solution to the Poisson equation for the pressure. Overset grids are
employed to effectively describe interfaces and physical boundaries for complex geometries. A
particular form of upwind dissipation is included in the scheme to ensure stability on overset grids
and traction boundaries. Divergence damping is added into the scheme to maintain small dilatations.
A Gustafsson, Kreiss, and Sundstrom (GKS) normal
mode analysis is performed on a model problem to verify the stability of this scheme with displacement and traction
boundary conditions. To showcase the accuracy and stability of the scheme, several numerical
experiments using known solutions for varying geometries are shown. The second main result is the development of a novel framework for the construction of high-order accurate finite difference schemes for the wave equation, Maxwell's equations, and incompressible linear elasticity. The framework consists of a hierarchy in which high-order accurate approximations are formed from lower-order ones.Each level within this hierarchy is constructed using a second-order accurate finite difference scheme. To obtain higher-order accuracy, the second-order scheme is augmented with corrections which are found according to a modified-equation approach. In addition, higher-order accurate discrete boundary and interface conditions are constructed as part of this hierarchy. The overall schemes use only three time levels and are CFL-one stable. Application of von Neumann analysis shows the stability of these high-order accurate
hierarchical schemes and numerical experiments verify the predicted stability and accuracy.Ph
Semantically enabling clinical decision support recommendations
No full textBackground
Clinical decision support systems have been widely deployed to guide healthcare decisions on patient diagnosis, treatment choices, and patient management through evidence-based recommendations. These recommendations are typically derived from clinical practice guidelines created by clinical specialties or healthcare organizations. Although there have been many different technical approaches to encoding guideline recommendations into decision support systems, much of the previous work has not focused on enabling system generated recommendations through the formalization of changes in a guideline, the provenance of a recommendation, and applicability of the evidence. Prior work indicates that healthcare providers may not find that guideline-derived recommendations always meet their needs for reasons such as lack of relevance, transparency, time pressure, and applicability to their clinical practice.
Results
We introduce several semantic techniques that model diseases based on clinical practice guidelines, provenance of the guidelines, and the study cohorts they are based on to enhance the capabilities of clinical decision support systems. We have explored ways to enable clinical decision support systems with semantic technologies that can represent and link to details in related items from the scientific literature and quickly adapt to changing information from the guidelines, identifying gaps, and supporting personalized explanations. Previous semantics-driven clinical decision systems have limited support in all these aspects, and we present the ontologies and semantic web based software tools in three distinct areas that are unified using a standard set of ontologies and a custom-built knowledge graph framework:
(i) guideline modeling to characterize diseases,
(ii) guideline provenance to attach evidence to treatment decisions from authoritative sources, and
(iii) study cohort modeling to identify relevant research publications for complicated patients.
Conclusions
We have enhanced existing, evidence-based knowledge by developing ontologies and software that enables clinicians to conveniently access updates to and provenance of guidelines, as well as gather additional information from research studies applicable to their patients’ unique circumstances. Our software solutions leverage many well-used existing biomedical ontologies and build upon decades of knowledge representation and reasoning work, leading to explainable results