DSpace@RPI (Rensselaer Polytechnic Institute)
Not a member yet
6809 research outputs found
Sort by
Embedding fairness and ethics in collective decision-making
No full textAugust 2023School of ScienceCollective decision-making is the problem where we have to aggregate individual preferences to collectively make a choice. Voting is one of the most commonly studied methods of making collective decisions. The task of collective decision-making can generally be divided into two parts: 1. preference learning and 2. preference aggregation. In the preference aggregation domain, judging the quality of voting rules in terms of well-defined properties and paradoxical behaviors is an important topic. Additionally, with the recent growth of algorithmic decision-making, concerns regarding fairness and ethics are being discussed more and more in the voting domain. For example, how can we explicitly guarantee decisions that are fair towards minority groups? How can we learn preferences in an ethical domain? Can we embed ethical properties in the preference aggregation method? This dissertation aims at answering these questions. With the advent of machine learning techniques, particularly in preference learning and similar use cases, we ask the question, how can we better use these techniques to help make better collective decisions? The first part of this work focuses on introducing a new notion of group fairness in voting. First, we explicitly consider the group identity of agents in order to make collective decisions that are fair towards the minority. This explicit consideration of agent identities is opposed to the commonly considered anonymity property, which ensures all agents are treated the same. Next, we analyze group fairness guarantees for existing voting rules, focusing on economic efficiency, and we see a trade-off between economic efficiency and group fairness. Then, we focus on designing voting rules that are both fair and efficient. For this, we develop machine learning-based techniques for automatically designing new voting rules that can achieve different levels of trade-offs between fairness and efficiency. Next, we work on learning agent preferences in a moral dilemma. Since the ethical domain is considered high-stakes, we explore learning explainable models to represent agent preferences. We collect a new dataset of agent preferences in moral dilemmas and experiment with learning heuristic models like lexicographic preferences and efficiently aggregating individual models to represent a social-level preference model for moral dilemmas. Lastly, we present a slightly different but related work on verification algorithms for no-show paradoxes under popular voting rules. Here, we develop different algorithms based on integer linear programs and heuristic-based searches that let us algorithmically verify whether a group no-show paradox is possible under a specific voting rule given a voting profile.Ph
Hybrid systems and the role of the linker molecule in charge transfer heterostructures : a first principles study
No full textDecember 2022School of ScienceHybrid systems are composed of materials like nanorods and sheets that have reduced dimensionality. These hybrid systems have the potential to revolutionize the way we harvest energy and design systems at the nanoscale. One such example---colloidal quantum dots---are a 0D material, and show remarkable properties useful for photovoltaics and photocatalysis. The ligand molecules that tether these QDs to another surface possess a role that is not well understood. In this thesis, first principles calculations are employed to study these hybrid systems. First, density functional theory is used to understand how a model linker molecule (cysteine) chemically links the QD to the surface. We find configurations for cysteine in various protonation states linked to CdSe(100) and CdSe(001) via a Monte Carlo simulation. Next, we investigate what effect this linker molecule has on the charge transfer between the QD and the surface using the novel three-way-heterostructure from the previous work. We find that the linker molecule, when interacting with both sides, undergoes an intramolecular charge transfer, and the resulting system has greatly enhanced charge transfer based on Marcus Theory. Then, a dynamical simulation of this heterostucture was performed using time-dependent DFT (TD-DFT) in order to explore how the ligand affect the charge transfer at the ultra-fast timescale. We find that the ligand molecule and increased ionic temperature, enhance the charge transfer at this timescale, and that electrons are preferentially transferred from CdS to MoS2 as band alignment would predict. Finally, the study of hybrid systems is brought to a twisted bilayer of 2-dimensional h-BN, where the possible Moire patterns are analyzed and enumerated, and then a study of intercalated transition metal defects is conducted.Ph
Learning based model correction and data imputation algorithms for rotorcraft systems
No full textAugust 2023School of EngineeringThe future of vertical takeoff and landing (VTOL) vehicles encompasses numerous thrilling and innovative designs, including electric multirotor vehicles and high-speed compound helicopters. Electric multirotor vehicles are expected to be utilized for various purposes such as human transportation ("air taxis"), package delivery, and surveillance, among others. On the other hand, the development of high-speed compound helicopters aims to cater to military missions, including aerial scout missions such as assault and reconnaissance, as well as troop transport. The lack of high-fidelity flight simulation models for rotorcraft poses a significant challenge in understanding complex systems and limits the efficiency and cost-effectiveness of rotorcraft development and evaluation processes. Secondly, the impact of sensor data loss on autonomous rotorcraft operations raises concerns about compromised behavior and safety mid-flight. To tackle these problems, this research presents a novel learning-based framework for model correction and data imputation tailored to different rotorcraft platforms. Leveraging machine learning approaches that have shown promise in model correction, multifidelity modeling, and data imputation, this framework aims to enhance the development of accurate simulation models for rotorcraft and provide solutions for sensor data loss. By employing machine learning techniques, the proposed solution seeks to improve the fidelity of simulation models, reduce the reliance on costly experiments or high-fidelity simulations, and ensure reliable decision-making and conditional assessment during autonomous flights. This body of work explores the development of a model correction algorithm for a low-order model of a compound helicopter with targeted high-fidelity data at specific areas of interest using machine learning approaches. The method is applied to a low-fidelity comprehensive trim analysis of a compound helicopter with three degrees of control redundancy: main rotor speed, auxiliary thrust, and stabilator setting. The final low-fidelity correction model applies small changes to the power requirement and main rotor trim control predictions to match the high-fidelity data. To reduce the computational time, and labor cost of querying the high-fidelity data, the algorithm prioritizes data acquisition by iteratively selecting the data where the error is expected to exceed the model tolerance by the greatest margin. Next, this work presents a novel methodology for identifying nonlinear corrections to improve the accuracy of a physics-based simulation model of a hexacopter using flight test data. The nonlinear corrections are identified by analyzing correlations between different flight variables and utilizing a filtered dataset with a high normalized correlation. The regularized version of partial least squares is applied for identifying the correction terms. For imputing missing sensor data in multicopters, which enables enhanced safety and reliability, a deep learning-based framework is developed by leveraging the multitude of sensors on these aerial vehicles. The proposed approach is based on two deep learning techniques, namely Autoencoders (AE) and Long Short-Term Memory (LSTM) networks. The effectiveness of this approach is evaluated using flight test data from a 2.5 kg hexacopter, and three different scenarios of missing data are considered.To validate the performance of the proposed approach, it is compared against two commonly used imputation techniques: k-Nearest Neighbor (KNN) imputation and Random Forest imputation.Ph
Facilitating Reuse of Mental Health Questionnaires via Knowledge Graphs
No full textQuestionnaires are one of the most common instrument types for screening patients for mental disorders. They are composed of items whose answers are typically scored to determine the elevation on a specified dimension, and hence the statistical probabilities associated with the corresponding disorder or diagnosis. The Patient Health Questionnaire (PHQ-9) and the Generalized Anxiety Disorder (GAD-7) questionnaire, for instance, measure levels of depression and anxiety respectively, and can be used to support diagnosis of depression and generalized anxiety disorder. Some questionnaires are multidimensional, such as the Revised Children's Anxiety and Depression Scale (RCADS), and can thereby estimate elevations on multiple dimensions that underlie a variety of disorders. Mental health screening questionnaires are designed so that each item assesses specific symptoms whose pattern of co-occurence (often organized in a subscale) allows estimation of how likely such symptoms would occur in the absence of the disorder whose symptoms the items represent. Questionnaire users typically estimate how likely a set of co-occuring symptoms would be (i.e., a score) in the general population as a strategy to estimate the likelihood that the respondent has a disorder warranting mental health services.
The RCADS is a 47-item, youth self-report questionnaire with subscales (separation anxiety disorder, social phobia, generalized anxiety disorder, panic disorder, obsessive compulsive disorder, and major depressive disorder). It also yields a Total Anxiety Scale (sum of the 5 anxiety subscales) and a Total Internalizing Scale (sum of all 6 subscales). Items are rated on a 4-point Likert-scale from 0 ("never") to 3 ("always"). A Parent Version (RCADS-P) similarly assesses parent report of a youth’s symptoms across the same six subscales. Brief versions of the RCADS questionnaires are available as well (RCADS-25), yielding only three scores: Total Anxiety, Total Depression, and Total Anxiety and Depression. RCADS questionnaires have been translated to 19 languages.
Recently there has been increased interest in supporting widespread adoption of common measures in mental health to support estimation and measurement of clinical dimensions across settings, contexts, and nations. However, the current measurement architecture in mental health is essentially based on a text-only document representation of each questionnaire, with limited knowledge of how they were created, how they relate to other questionnaires, how items relate to symptoms, which in turn relate to disorders, how short and long versions are related, etc. The result is significant constraints on the types of use cases that can be supported, with especially limited support for such pursuits as shortening the number of items, translations to new languages, reuse of items in new questionnaires, and even the combination of items from different questionnaires.
We present our progress towards tackling these challenges. Our solution is composed of: (1) a modeling of mental health symptoms, scales, disorders, and their relationships as an ontology; (2) the representation of questionnaire instruments as a knowledge graph, using standardized terminology; and (3) a software infrastructure for operationalizing the management and distributions of semantic questionnaires (Semantic Instrument Repository - SIR). Using the RCADS questionnaires as a use case, we encode their (sub)scales in an ontology, reusing existing terminology from relevant sources. We expand our base Human-Aware Science Ontology (HAScO) to include questionnaire structure, and propose a new ontology for encoding and aligning mental health terminology, such as symptoms, scales, and disorders. SIR supports authoring, curation, and dissemination of questionnaires, their elements, and relationships between these elements, thus allowing questionnaires to contain mental health semantics
Design of a cold moderator for total cross section measurements of moderator materials at sub-thermal energies
No full textDecember 2022School of EngineeringIn order to extend the neutron cross section measurement capability to below 0.001 eV, the sub-thermal neutron flux of the existing enhanced thermal target needed to be improved at the Rensselaer Polytechnic Institute (RPI) linear accelerator (LINAC). To meet this need, a novel polyethylene based cold moderator capability was designed and constructed. When coupled to the enhanced thermal target, the neutron flux was increased by up to a factor of 8 below 20 meV after background subtraction. A new neutron producing target was also designed in order to handle higher electron beam intensities that would be present with the new proposed upgrade to the RPI LINAC.This cold moderator capability was used for transmission measurements of various moderating materials, including yttrium hydride at a hydrogen concentrations of 1.85 and 1.68, polyethylene, polystyrene and Plexiglas from 0.0005 - 3 eV. These measurements were compared directly to thermal scattering library (TSL) evaluations and existing experiments where applicable. Generally good agreement was found, but some discrepancies were noticed and will be discussed. The polyethylene measurements validated the use of the cold moderator system and its method, while extending the measured cross section to 0.0005 eV. For yttrium hydride, the measured range was extended above 0.8 eV and below 0.05 eV, representing the first measurements to encompass the entire thermal energy range. The Plexiglas measurements represented the first where the material was well known and extended the measured cross section above 1 eV and below 0.002 eV. These measurements are the first total thermal cross section measurements for polystyrene.Ph
Narrativized re-performances of gameplay: witch-players and methodological resistance in game studies
No full textAugust 2020School of Humanities, Arts, and Social SciencesThe primary purpose of this dissertation is to develop a hybrid methodology for textual studies of videogames. At its core is a reconceptualization of gameplay, a term with industry origins that refers to players’ engagements with videogames, but that has served to maintain numerous, gendered binaries prevalent in game studies scholarship. Against these trends, I argue that the textual meanings of gameplay emerge from the dynamic, assembled agencies of videogame technologies, designed gameworlds, and player subjectivities. As a basis for comprehending gameplay, I examine research that characterizes this composite activity as cyborgian, alongside feminist research in which cyborgs exemplify tensions between structural oppressions and agentic subjectivities. However, by explicating subversive feminine gameplay performances that operate simultaneously within and against videogames, I shift cyborgian sensibilities to posit the figure of the witch-player. This formulation at once captures the uptake of witches in contemporary popular culture that challenge gender and sexual norms, while also interrogating historical roots in misogyny and racism that have served the spread of capitalism and settler-colonialism. Further, it intervenes in game studies research by foregrounding feminine gameplay performances. To read gameplay, I construct a methodology that combines textual analysis, autoethnography, and assemblage theories. Demonstrating its use, I chronicle my resistant, feminine gameplay in various horror videogames, providing a condensed genre study throughout these analyses.Ph
Investigation and projection of near- and far-term dynamic glazing systems for dynamic bioclimatic façades
No full textDecember 2015School of ArchitectureWindows have a significant impact on building energy use but additionally influence occupant comfort and the architectural aesthetic, creating a complex, multivariate, and interdisciplinary challenge. Contemporary buildings continue to pursue highly glazed envelopes despite conflicting issues with human comfort, health and energy use and have necessitated research into advanced windows for energy efficient buildings. While existing advanced window technologies have made incremental progress towards greater energy savings and commercialization, they remain limited in their comprehensive treatment of interdependent concerns. Emerging material possibilities not only point towards new physical phenomena for achieving transparency modulation, but demand a broader reinterpretation of the performance criteria and end goals for advanced glazing systems. If the performance criteria and research methodologies guiding glazing technology development are complicit in producing low-impact products that falter in the market place due to single-function value and limited socio-cognitive performance, then a broad trans-disciplinary research methodology operating under revised and comprehensive performance criteria will support the development of technologies with multiple value propositions capable of the diverse and ubiquitous implementation needed for a high-impact. In order to guide the development of ubiquitous dynamic façade technologies with multiple value propositions this research defines an expanded set of performance criteria and investigates three interrelated embodiments of dynamic façade pathways to meet the criteria: the electropolymeric dynamic daylighting system, a novel graphene oxide optical modulator, and functionalized graphene oxide for energy harvesting through solar water oxidation. In addition to highlighting an improved model of trans-disciplinary collaboration, this research critically expands the longstanding definition of the dynamic glazing problem, and advances co-modeling techniques, novel light modulating materials, and innovative energy harvesting materials. Together these represent a pathway for the development of multifunctional dynamic façades that are capable of satisfying criteria for enhanced human comfort, energy efficiency, and energy conversion.Ph
Understanding the limits of AI coding
No full textIn the 9 December 2022 issue, the Research Article “Competition-level code generation with AlphaCode” (Y. Li et al., p. 1092) and the accompanying Perspective, “AlphaCode and ‘datadriven’ programming” (J. Z. Kolter, p. 1056) describe an artificial intelligence (AI)–based system for generating code. The authors explain that the system can be used for small coding problems, such as tests for computing students, and that they are far from being useful in computing applications that include millions of lines of code, such as word processing. As we enter an era of AI where tools like AlphaCode and chatGPT will change how tasks are performed, it is important to understand the boundaries of what they can and cannot do.
To make sure that code can be maintained and managed by other programmers, human developers use mnemonic variable names and embed explanatory comments. Understanding, debugging, and extending code written by other humans remains a formidable challenge—perhaps even more difficult than producing the code in the first place. In addition, many techniques are used for validation and verification, and code used in mission-critical applications, such as airline flight systems, goes through substantial quality assurance testing. AI models have yet to address the challenges of maintaining code, ensuring that users can decipher it, and subjecting programs to safety protocols.
Understanding and evaluating the limits of these techniques is crucial before they are put into real-world use. Some testing of capabilities has been applied to language generation tools (1, 2), but AI coding remains a nascent field. The Technology Policy Committee of the Association for Computing Machinery recommends more investment in transparency and accountability for AI algorithms (3). The promise of systems like AlphaCode must be carefully balanced against the risks of their use. The interaction between AI code-generation systems and human programmers must be resolved before such systems can become an integral part of the future of computing
Kap is a neuronal organelle adaptor for KIF3AB and KIF3AC
No full textDecember 2022School of ScienceNeurons are the fundamental unit of the nervous system, their function being to receive and propagate electrochemical signals. Due to this role, neurons can be exceptionally large cells with specialized subdomains known as dendrites and axons. To perform their signal reception and propagation functions, dendrites and axons contain unique complements of proteins which must be transported and delivered to their required destination. These proteins are packaged into membrane-bound organelles which molecular motors, such as kinesins, bind and transport. Kinesin-driven organelle transport is crucial for neuron development and maintenance, yet the mechanisms by which kinesins recognize and bind their specific organelle cargoes remain poorly defined. The neuronal function and specific organelle adaptors of heterodimeric Kinesin-2 family members KIF3AB and KIF3AC remain unknown. I developed a novel microscopy-based assay to define protein–protein interactions in intact neurons. The experiments revealed that KIF3AB and KIF3AC both bind kinesin-associated protein (KAP) and that these interactions are mediated by the distal C-terminal tail regions and not the coiled-coil domain. I used live-cell imaging in cultured hippocampal neurons to define the localization and trafficking parameters of KIF3AB and KIF3AC organelle populations. KIF3AB/KAP and KIF3AC/KAP bind the same organelle populations, and I defined their transport parameters in axons and dendrites. The results also show that ~12% of KIF3 organelles contain the RNA binding protein, adenomatous polyposis coli. These data point towards a model in which KIF3AB and KIF3AC use KAP as their neuronal organelle adaptor and that these kinesins mediate transport of a range of organelles.
In a separate project, I described a novel strategy to allow for consistent visualization of kinesin-bound organelles in live mammalian neurons. Previous attempts to label kinesin-bound organelles in live cells utilized expression of fluorophore-fused full-length kinesins. This strategy results in a diffuse, cytosolic expression pattern which obscures labeled organelles. This large fraction of cytosolic expression is hypothesized to be due to unbound autoinhibited motor. Therefore, reduction of this cytosolic, unbound kinesin pool is crucial for visualization of organelle-bound kinesins.
I describe two strategies that improve visualization of vesicle-bound kinesins. The first is a truncation strategy where only the organelle-binding tail domain of kinesins are expressed. Truncated kinesins only expressing organelle-binding tail domains are unable to form an autoinhibited conformation, increasing the amount of exogenously expressed protein available to bind to organelles. The second is a transcriptional control technique where constructs are designed with a nuclear localization signal and a zinc finger domain that acts as a plasmid-specific transcription repressor. Upon translation, any unbound kinesin tail is targeted to the nucleus where it represses its own transcription. Using these strategies drastically improves the imaging conditions for organelle-bound kinesins in live hippocampal neurons.Ph
Interactions of the mediodorsal thalamus and the prefrontal cortex support cognitive flexibility
No full textAugust 2023School of Humanities, Arts, and Social SciencesCognitive control has long been a central problem of neuroscience. Recent experimental findings suggest that several subcortical regions are implicated in mediating brain function which is a significant shift from the cortico-centric paradigm. One of them, the mediodorsal thalamic nucleus, has been shown to influence prefrontal function. Yet, understanding of the function performed by this thalamocortical circuit and its place in the cognitive control system is limited and requires further investigation. A neural network model of mediodorsal-prefrontal function is used in a series of computational experiments to investigate the link between the topological properties of the brain connectivity and the function of the mediodorsal thalamus. The results of the present research are twofold. It began by further developing network theory methodology utilizing artificial neural networks. Accordingly, the encoding of the functional content of a brain region is facilitated by the connectivity and neurochemical properties of its neurons. Second, this methodology is used in the in silico investigation of recent phenomena related to the function of the mediodorsal thalamus. Computational modeling suggests that the function of the mediodorsal thalamus is primarily associated with the neuromodulatory properties of its efferent projections. Furthermore, computational experiments show how this function contributes to cognitive flexibility. Additionally, computational experiments suggest a link between suppression in the activity of the mediodorsal thalamus and disruption of prefrontal activity, one of the most common symptoms of schizophrenia.Ph