21090 research outputs found
Sort by
The Interplay of Information Theory and Deep Learning: Frameworks to Improve Deep Learning Efficiency and Accuracy
No full textThe intersection of information theory (IT) and machine learning (ML) represents a promising, yet relatively under-explored, frontier with significant potential for innovation. Despite the clear benefits of combining these fields, progress has been limited by two main challenges: (i) the highly specialized nature of IT and ML, which creates a barrier to cross-disciplinary expertise, and (ii) the computational complexity involved in applying information-theoretic concepts to large-scale ML problems.
This dissertation seeks to overcome these challenges and explore the rich possibilities at the intersection of IT and ML. By leveraging powerful tools and concepts from IT, we aim to uncover novel insights and develop innovative ML algorithms. Given that deep neural networks (DNNs) form the backbone of modern ML models, the integration of IT principles into ML requires a focus on optimizing the training and performance of DNNs
using information-theoretic frameworks.
While DNNs have a broad range of applications, this thesis narrows its focus to two key areas: classification and generative DNNs. The objective is to harness IT principles to enhance the performance of these models.
• Classification DNNs. For classification DNNs, this dissertation targets improvements in three critical areas:
(i) Improving classification accuracy. The performance of classification DNNs is traditionally measured by classification accuracy, but we argue that conventional error metrics are insufficient for capturing a model’s true performance. By introducing the concepts of conditional mutual information (CMI) and normalized conditional mutual information (NCMI), we propose a new metric for evaluating DNNs. The CMI measures intra-class
concentration, while the ratio of CMI to NCMI reflects inter-class separation. We then modify the standard loss function in deep learning (DL) framework to minimize the standard cross entropy function subject to an NCMI constraint, yielding CMI constrained deep
learning (CMIC-DL). Then, via extensive experiment results, we show that DNNs trained within CMIC-DL achieves a higher classification accuracy compared to the state-of-the-art models trained within the standard DL and other loss functions in the literature.
(ii) Enhancing distributed learning accuracy. In the context of distributed learning, particularly federated learning (FL), we tackle the challenge of class imbalance using informationtheoretic concepts to improve the accuracy of the shared global model. To this end, we introduce new information-theoretic quantities into FL and propose a modified loss function based on these principles. This leads to the development of a federated learning framework, Fed-IT, which enhances the classification accuracy of models trained in distributed environments.
(iii) Reduce the size and training/inference complexity. We introduce coded deep learning (CDL), a novel framework aimed at reducing the computational and storage complexity of classification DNNs. CDL achieves this by compressing model weights and activations through probabilistic quantization. Both forward and backward passes during training are performed using quantized weights and activations, significantly reducing floating-point operations and computational overhead. Furthermore, CDL imposes entropy constraints on weights and activations, ensuring compressibility at every stage of training, which also reduces communication costs in parallel computing environments. This leads to models that are more efficient in both training and inference, with lower storage and computational requirements.
• Generative DNNs. For generative DNNs, this dissertation focuses on diffusion models and their application to solving inverse problems. Inverse problems are common in fields like medical imaging, signal processing, and physics, where the goal is to recover an underlying cause from corrupted or incomplete observations. These problems are often ill-posed, with multiple possible solutions or high sensitivity to small changes in the data. In this dissertation, we enhance the performance of diffusion models by incorporating probabilistic principles, making them more effective at capturing the posterior distribution of the underlying causes in inverse problems. This approach improves the model’s ability to accurately reconstruct signals and provides more reliable solutions in challenging inverse problem scenarios.
Overall, this dissertation demonstrates the powerful synergy between IT and ML, showcasing novel methods that improve the accuracy and efficiency of both classification and generative DNNs. By addressing key challenges in training and optimization, this work lays the foundation for future research at the intersection of these two fields
LiDAR-Driven Calibration of Microscopic Traffic Simulation for Balancing Operational Efficiency and Prediction of Traffic Conflicts
No full textMicroscopic traffic simulation is a proactive tool for road safety assessment, offering an alternative to traditional crash data analysis. Microsimulation models, such as PTV VISSIM, replicate traffic scenarios and conflicts under various conditions, thereby aiding in the assessment of driving behavior and traffic management strategies. When integrated with tools like the Surrogate Safety Assessment Model (SSAM), these models estimate potential conflicts. Research often focuses on calibrating these models based on traffic operation metrics, such as speed and travel time, while neglecting safety performance parameters.
This thesis investigates the effects of calibrating microsimulation models for both operational metrics including travel time and speed, and safety metrics including traffic conflicts and Post Encroachment Time (PET) distribution, using LiDAR sensor data. The calibration process involves three phases: performance calibration, performance and safety calibration, and only safety calibration. The results show that incorporating safety-focused parameters enhances the model's ability to replicate observed conflict patterns. The study highlights the trade-offs between operational efficiency and safety, with adjustments to parameters like standstill distance improving safety outcomes without significantly compromising operational metrics. Furthermore, there is a substantial difference in the calibrated minimum distance headway for the safety model, highlighting the trade-off between operational efficiency and safety. While the operational calibration focuses on optimizing flow, the safety calibration prioritizes realistic conflict simulation, even at the cost of reduced flow efficiency. The research emphasizes the importance of accurately simulating real-world driver behavior through adjustments to parameters like the probability and duration of temporary lack of attention
A Survival-Driven Machine Learning Framework for Donor-Recipient Matching in Liver Transplantation: Predictive Ranking and Optimal Donor Profiling
No full textLiver transplantation is a life-saving treatment for patients with end-stage liver disease. However, donor organ scarcity and patient heterogeneity make finding the optimal donor-recipient matching a persistent challenge. Existing models and clinical scores are shown to be ineffective for large national datasets such as the United Network for Organ Sharing (UNOS). In this study, I present a comprehensive machine-learning-based approach to predict posttransplant survival probabilities at discrete clinical important time points and to derive a ranking score for donor-recipient compatibility. Furthermore, I developed a recipient-specific "optimal donor profile," enabling clinicians to quickly compare waiting-list patients to their ideal standard, streamlining allocation decisions. Empirical results demonstrate that my score’s discriminative performance outperforms traditional methods while maintaining clinical interpretability. I further validate that the top compatibility list generated by our proposed scoring method is non-trivial, demonstrating statistically significant differences from the list produced by the traditional approach. By integrating these advances into a cohesive framework, our approach supports more nuanced donor-recipient matching and facilitates practical decision-making in real-world clinical settings
Toward Adaptive and User-Centered Intelligent Vehicles: AI Models with Granular Classifications for Risk Detection, Cognitive Workload, and User Preferences
No full textAs artificial intelligence (AI) increasingly integrates into our transportation systems, intelligent vehicles have emerged as research topics. Many advancements aim to enhance both the safety and comfort of drivers and the reliability of intelligent vehicles. The main focus of my research is addressing and responding to the varying states and needs of drivers, which is essential for improving driver-vehicle interactions through user-centered design. To contribute to this evolving field, this thesis explores the use of physiological signals and eye-tracking data to decode user states, perceptions, and intentions. While existing studies mostly rely on binary classification models, these approaches are limited in capturing the full spectrum of user states and needs. Addressing this gap, my research focuses on developing AI-driven models with more granular classifications for cognitive workload, risk severity levels, and user preferences for self-driving behaviours. This thesis is structured into three core domains: collision risk detection, cognitive workload estimation, and perception of user preferences for self-driving behaviours. By integrating AI techniques with multi-modal physiological data, my studies develop ML (Machine Learning) models for the domains introduced above and achieve high performance of the ML models. Feature analytical techniques are employed to enhance model interpretability for a better understanding of features and to improve the model performance. These findings pave the way for a new paradigm of intelligent vehicles that are not only more adaptive but also more aligned with user needs and preferences. This research lays the groundwork for the future development of user-centered intelligent companion systems in vehicles, where adaptive, perceptive, and interactive vehicles can better meet the complex demands of their users
TECTONIC EVOLUTION OF THE BAIE VERTE MARGIN, NEWFOUNDLAND
No full textLocated along the Early Paleozoic Laurentian continental margin in Newfoundland, the Baie Verte Margin tectonistratigraphy and tectono-metamorphic evolution have been controversial for decades. Here, the results of a detailed field, petrological and geochronological study are presented, where Baie Verte Margin is subdivided into three tectono-metamorphic units separated by tectonic contacts: the East Pond Metamorphic Suite (EPMS) basement, EPMS cover, and the Fleur de Lys Supergroup (FdLS). Each unit exhibits a distinct metamorphic and structural evolution recorded during the subduction, exhumation, and post-collisional history of this ancient margin. The combination of thermodynamic modelling, petrochronology, and structural analysis provided insights into the P-T-t-d paths of the studied units, allowing a better understanding of their role during the evolution of the Taconic subduction system. High-pressure (HP) to ultra-high-pressure (UHP) conditions were reached between 483 and 475 Ma during the D1 phase, with the EPMS cover recording eclogite-facies metamorphism at ~2.8 GPa and 620°C. Subsequent decompression resulted in a β-shaped pressure-temperature-time (P-T-t) path, with near-isothermal decompression to ~2 GPa and heating to 860°C during exhumation. A multi-stage exhumation model is proposed for the EPMS eclogites: 1) buoyant rise through a low-density mantle wedge and 2) subsequent ascent at shallower crustal levels, facilitated by external tectonic forces and slab break-off, as evidenced by Late Taconic magmatism.
While the EPMS cover re-equilibrated at UHP conditions, the EPMS basement and FdLS experienced decompression and Barrovian metamorphism during late-D1, indicating decoupling of the units during this stage. Coupling between the units occurred along a D2 shear zone during retrograde metamorphism, spanning 475–452 Ma.
Two exhumation scenarios are proposed to explain the tectonic evolution of the margin: (i) Following late D1 detachment, the EPMS basement and FdLS were exhumed to crustal levels while the EPMS cover was subducted deeper into the mantle. Tectonic extrusion along D2 shear zones, potentially aided by melt weakening, then emplaced the EPMS cover between the two units. (ii) Alternatively, sequential detachment occurred from the top to the bottom of the slab, resulting in deeper subduction of lower units, followed by their exhumation through back-folding and crustal wedge thrusting.
The Silurian F3 folding deforms both D1-2 structures in each unit and the D2 shear zones that bound them, suggesting that the continental wedges, which recorded different tectono-metamorphic paths after early D1, were juxtaposed before the onset of deformation associated with the Salinic Orogeny. Later deformation phases, D4 and D5, are probably related to tectono-metamorphic activity related to the Acadian and Neo-Acadian orogenies.
This research improves our understanding of the dynamic tectono-metamorphic evolution of the Baie Verte Margin, emphasizing the role of fluids, thermal perturbations, and deformation in driving metamorphic reactions, and exhumation. The findings contribute to understanding the mechanisms controlling HP-UHP terrain evolution in subduction zones and highlight the complex interactions between subduction, exhumation, collision, and magmatism throughout the Taconic orogeny
Improving Pyrometry of Advanced High Strength Steels During Intercritical Annealing
No full textAdvanced high strength steels (AHSS) play a prominent role in the automotive industry due to their unique combinations of strength and ductility—essential material properties for manufacturing lightweight vehicles that meet global regulations for fuel economy, vehicle emissions, and passenger safety. Precise thermal control during intercritical annealing is crucial to achieving the mechanical properties of AHSS. Unfortunately, steel manufacturers report unacceptably high rejection rates for AHSS due to substandard mechanical properties. This problem is often attributed to temperature excursions during intercritical annealing of the AHSS, which can be caused by errors in the pyrometrically-inferred temperatures used to control the furnaces. Measuring the temperature of the steel strip through pyrometry requires detailed knowledge of the spectral emissivity of the steel strip, which is imperfectly known since it varies with wavelength, direction, temperature, surface roughness, and oxidation, the latter depending on alloy composition and processing conditions. Therefore, wavelength-dependent variations in the spectral emissivity of AHSS lead to errors in pyrometry measurements during intercritical annealing, which in turn affect the mechanical properties of the steel.
In pyrometry, a temperature estimate is typically obtained using spectral irradiance measured from the steel strip in combination with an emissivity compensation algorithm that assumes or prescribes an underlying emissivity relationship based on the number of detection wavelengths. While various emissivity compensation approaches have been developed for mitigating pyrometer errors caused by uncertain spectral emissivity, none have been specifically designed for AHSS. Hence, these methods do not adequately capture how the evolving surface state of annealing AHSS affects the spectral emissivity and thus the resulting pyrometrically-inferred measurements. Further, existing pyrometry algorithms only provide a point estimate of the surface temperature, making it impossible to assess the reliability of the estimates. To address these challenges, this dissertation aims to develop robust pyrometry methods for AHSS that provide accurate and precise temperature estimates with its associated uncertainties.
This thesis starts by presenting an empirical approach for modelling the spectral emissivity of advanced high strength steel based on response surface methodology (RSM). Using ex situ measurements on annealed dual-phase AHSS samples (DP980), variation in the spectral emissivity with respect to dew point, alloy composition, pre-annealed surface state, and wavelength is analyzed using full factorial designs. The significant main and interaction effects were found to vary across the spectral range, with the ratio of alloy components and pre-annealed surface state dominating at shorter and longer wavelengths, respectively. Furthermore, the factorial design of experiments was used to develop a novel multivariate emissivity model capturing the effects of dew point, alloy composition, pre-annealed surface state, and wavelength on the emissivity of dual-phase AHSS.
To extend the investigation to the high-temperature spectral emissivity variations during intercritical annealing, a laboratory-scale annealing simulator representative of the industrial furnace conditions was designed and fabricated to enable in situ pyrometry and emissivity measurements on AHSS samples. With lab-scale experiments that simulate the annealing and temperature control process of an industrial continuous galvanizing line (CGL), the in situ effect of process parameters, such as chemical composition, annealing temperature, and atmosphere dew point, on the radiative properties of AHSS was observed and modelled. Using in situ measurements from DP980 alloys heated in the annealing simulator, the suitability of existing pyrometry models for accurately predicting the surface temperature of AHSS was examined. By comparing the performance of several dual-wavelength (ratio) and multi-wavelength pyrometry algorithms, it was observed that both methods over-predict the surface temperature; however, the predictions of the multi-wavelength algorithm were generally superior.
Given that the spectral emissivity varies significantly with the surface properties of the AHSS coil and further evolves with surface oxidation during annealing, this thesis also investigated the effect of the annealing atmosphere on the radiative properties of dual phase AHSS (DP980) using in situ spectral emissivity measurements from samples annealed within a reducing N2-H2 atmosphere at dew points ranging from −45°C to +10°C, with an annealing schedule similar to that of industrial continuous galvanizing lines (CGLs). The analysis further explored the effect of variations in oxidation kinetics by comparing the radiative properties of DP980 replicates annealed with the same heating schedule and dew point. It was also discovered that low-temperature oxidation of the native oxide of AHSS impacts the evolving spectral emissivity during intercritical annealing. In addition, a comparison of evolving spectral emissivity of dual-phase AHSS alloys against that of a standard EDDS grade IF steel showed that in situ emissivity measurements could potentially identify the formation of mixed/ternary oxides during annealing.
Furthermore, this thesis also evaluated the ability of ex situ measurements to capture temperature-dependent variations in emissivity due to electron mobility as described by the Drude model and the Hagen-Rubens theory. The ex situ radiative properties were found to underpredict the in situ spectral emissivity, especially at lower temperatures and at shorter wavelengths. This underprediction was also shown to significantly impact the accuracy of pyrometry estimates at those temperatures. The ex situ measurements were also used to validate the annealing simulator against an industry-standard galvanizing simulator.
Finally, with the ultimate objective of improving the robustness of pyrometrically-inferred temperature measurements, this dissertation presents a Bayesian pyrometry methodology in which all pyrometry variables including the measured spectral irradiance, spectral emissivity and inferred-temperature are expressed as random variables that obey probability density functions. Additional information about the spectral emissivity from ex situ characterization were first incorporated into the inference through maximum likelihood priors. The prior and measurement densities were propagated through Bayes’ equation to obtain the posterior densities. The posterior densities provide the pyrometrically-inferred temperature and, crucially, its associated uncertainties. Compared to standard pyrometry methods that provide a point estimate of surface temperature, the Bayesian framework infers the measurement uncertainty via the posterior probability density, which will allow galvanizers to better assess the reliability of the pyrometrically-inferred temperature. The credibility interval of the temperature posterior was further narrowed by defining a multivariate in situ emissivity prior conditioned on the annealing dew point and the equivalent blackbody temperature.
Overall, by investigating the behaviour of the spectral emissivity of AHSS during processing, specifically how it evolves with material properties and processing parameters, this thesis presents a comprehensive empirical approach to developing emissivity compensation algorithms that improves the accuracy and reliability of pyrometric temperature predictions on AHSS during annealing
Elevated Temperature Formability of Precipitation Hardenable Sheet Aluminum Alloys
No full textThis thesis investigates the elevated temperature forming behavior of two age-hardenable aluminum alloys, AA6013 and a developmental 7000-series alloy (denoted AA7xxx), intended for automotive applications. Warm (150°C - 250°C) and hot (>=300°C) forming conditions were considered as well as a range of starting tempers to ascertain the potential benefits of elevated temperature forming. A major outcome of this work was the development and detailed documentation of a physically motivated method to assess formability, termed the Enhanced Curvature Method (ECM). The ECM utilizes stereoscopic digital image correlation (DIC) to detect the onset of necking based on changes in surface curvature.
The ECM was validated at room temperature using AA5182-O sheet which exhibits a strong dynamic strain ageing effect that affects strain localization. The plane strain limit strains obtained using the ECM in Marciniak, Nakazima, and stretch-bend formability tests were consistent with two other formability characterization methods developed for room temperature, the ISO12004-2:2008 standard and the rate-dependent method developed by Volk and Hora (2011) termed the “linear best fit” (LBF) method. The average Marciniak plane strain limit strain for AA5182-O obtained using the ECM was 0.198, and in close agreement with the theoretical limit strain of 0.194 obtained from the instability model of Swift.
The ECM was subsequently applied to characterize the room and elevated temperature formability of 2 mm thick AA6013 and AA7xxx. Both alloys were tested using a 76.2 mm wide dog-bone specimen geometry in a Nakazima-type test, that produced near plane-strain conditions. For AA6013-T6, when characterized using the ECM, warm forming did not offer a meaningful improvement in formability with increasing temperature over the tested ranges. The plane strain forming limit at both 230°C and room temperature (RT) was measured as approximately 0.18 with comparable scatter bands.
To improve warm formability, AA6013 in a pre-aged (PA) temper was investigated. The PA temper was produced by solutionizing the material at 560°C for 10 minutes, water quenching, aging at 100°C for 4 hours then followed by a second water quench. The room temperature formability of the PA temper was significantly higher than that of the peak-aged temper with a major limit strain of 0.28. The elevated temperature formability of the PA temper, however, decreased relative to its room temperature formability. At 230°C, the major limit strain was reduced to 0.21 and attributed to aging occurring during the warm forming process. A significant outcome of the AA6013 work was the observation that a relatively short secondary aging cycle (less than 410 s) at 235°C after forming, combined with a pre-aged initial temper, could result in a stress-strain response similar to the baseline T6 temper, regardless of a subsequent paint-bake. This processing route offers a potential pathway to achieve T6 strength levels with a greater than 50% improvement in formability when comparing the RT PA limit strain of 0.28 to the baseline T6 limit strain of 0.18.
The effect of initial temper on the warm formability and aging response of AA7xxx was also investigated for three tempers: pre-aged (PA), peak-aged (T6), and over-aged (T76), all tested using a 76.2 mm dog-bone specimen. The PA temper exhibited the highest formability at room temperature, with a major true limit strain of 0.21. At 150°C, the limit strain for the PA temper increased to 0.24, representing a modest improvement. At higher temperatures (175°C and 200°C), the results were not as clear, with measured dome heights decreasing and strains relatively unchanged. Similar to the AA6013 PA temper, the formability gains at elevated temperature for the AA7xxx PA temper were modest. Warm forming at 200°C, followed by a paint-bake, resulted in both the PA and T6 tempers achieving ductility and strength levels similar to those of the as-received T76 temper.
The hot-stamping and die quenching (DQ) process was investigated as an alternative means to enhance the formability of AA7xxx. The DQ process resulted in a major limit strain of 0.76 in plane strain. Under an interrupted, isothermal (300°C) hot-stamping operation, a major limit strain of 0.53 in plane strain was observed. Both limit strains were significantly higher than the room temperature formability of 0.12 measured for the as-received T76 temper and that achieved using warm forming. Surface defects, however, such as orange peel, were observed for major strain levels in excess of approximately 0.45.
A major conclusion of this work is that while both AA6013 and AA7xxx alloys demonstrated only modest improvements in formability when warm formed in their as-received conditions, a better balance of strength and formability can be achieved by leveraging pre-aged starting tempers combined with post-stamping aging cycles. The die quenching process results clearly show that excellent formability of AA7xxx series alloys can be achieved with additional process complexity. The ECM, a key outcome of this work, offers an effective and robust method for quantifying formability at both room and elevated temperatures. Future work should explore faster warm forming processes to overcome the rapid aging effects observed in the PA tempers. Additionally, other aluminum alloys should be evaluated for warm forming and die quenching to identify alloys that may offer a greater positive response to elevated temperature forming
Dual Energy X-ray for Quantitative Analysis of areal Bone Mineral Density (aBMD) using a stacked Flat Panel Detector (FPD)
No full textThis study explores the application of a stacked flat panel detector (FPD) system for spectral dual-energy X-ray imaging aimed at quantifying areal bone mineral density (aBMD). The proposed technique enables single-exposure dual-energy acquisition using a conventional cone-beam X-ray source, eliminating the need for sequential exposures or specialized hardware. This method facilitates simultaneous high- and low-energy imaging by capturing energy-separated data directly at the detector level, thereby enhancing workflow efficiency and spatial alignment. Experimental validation using bone-equivalent phantoms demonstrates the system's potential for accurate aBMD estimation, with robustness against positional variation and tissue scatter. The findings suggest a promising pathway toward more accessible, portable, and cost-effective bone health assessment in clinical and point-of-care settings
Assessment of the viability of VISR: a mid-wavelength infrared (MWIR) multispectral imaging-based approach for remote flare CE quantification
No full textIn the upstream and downstream petrochemical industry, flaring is a common practice to dispose of the redundant by-products of crude oil extractions, including associated gas and highly reactive volatile organic compounds (HRVOCs), for a variety of regulatory, safety and economic purposes. The current consensus among government regulators and industry experts is that flaring typically occurs at a combustion efficiency (CE) higher than 98%. Recent studies, based on real-world observations and computational simulations, have called this into question. For example, a recent study, based on airborne sampling observations and unlit flare prevalence surveys, reported a mean flare CE of 91%, accounting for both inefficient flaring and unlit flares in the three largest basins in the US. This represents a five-fold increase in fugitive hydrocarbon emissions, primarily comprised of methane, compared to the presumed release rates, highlighting an opportunity for developing robust flare CE monitoring techniques to mitigate the adverse health and environmental impacts of the underappreciated flaring emissions.
This study presents a numerical assessment of video imaging spectro-radiometry (VISR), a mid-wavelength infrared (MWIR) multispectral technique, proposed for remote flare combustion efficiency quantification applications. The present analysis utilizes a series of computational fluid dynamics (CFD) simulations of a crosswind steam-assisted industrial flare, with a focus on three aspects: how approximations in the radiometric model impact the local “pixel-wise” CE, the validity of the approach for computing flare global CE using inferred local CE values, and the ability and limitations of VISR instrument to capture fuel that may be aerodynamically stripped from the combustion zone under crosswind conditions. The current assessment is conducted on a simplified version of the VISR instrument model using simulated broadband images generated over spectral bands adjusted for the key absorption features of three main by-products of flare combustion reaction: CO2 (4.2–4.4 µm), CO (4.5–4.9 µm), and CH4 (3.2–3.4 µm). The results highlight the accuracy of the proposed simplified VISR approach in predicting local CE within the VISR region-of-interest (ROI) yet flawed in terms of converting these values into a flare global CE, potentially leading to large biases from the actual flare CE. Ultimately, the VISR technique, due to reliance on mid-wavelength infrared imaging, is inherently incapable of quantifying unburned (cold) methane, allegedly stripped from the flare stack, without participating in the combustion process, due to the presence of a high crosswind over the flare stack, leading to a considerable overestimation of the true flare performance.
Keywords: Flares, Combustion Efficiency, Remote Sensing, Verification and Validation, Radiometric Measurements, Uncertainty Analysis, Bayesian Inference
Knowing Language: The Poetics of Epistemology in Jan Zwicky, Paul Muldoon, and Geoffrey Hill
No full textThis dissertation investigates the poetics of epistemology in the works of Jan Zwicky (1955-), Paul
Muldoon (1951-), and Geoffrey Hill (1932-2016), three contemporary poets who engage
epistemology by way of diverse interdisciplinary proxies. Principally, my dissertation shows that
in Zwicky, Muldoon, and Hill, poetry is not representative as a knowledge-producing discourse
but is instead meta-epistemological: as a cultural artefact of language that is exemplary for the
way it self-reflexively expresses epistemological themes (thinking, thought, knowledge, etc.),
poetry calls into question the stability of linguistic meaning by challenging the epistemological
assumptions and rhetorical commonplaces of other discourses on knowing—especially
philosophy (Zwicky), history (Muldoon), and theology (Hill)—discourses whose epistemological
foundations are based on a language-as-knowledge-producing model.
For Zwicky, the paradigm of “lyric philosophy” informs and is informed by the poem’s
capacity as a phenomenological gestalt, where poetry’s knowing occurs in a matrix of linguistic
resonances. Gestalt insight in Zwicky’s work relies for its rhetorical force on the lyric integration
of linguistic elements rather than on language’s formal logical procedures. Muldoon’s framework
for poetic knowing—and not-knowing, and un-knowing—is the result of an aesthetics of
encyclopedic reference, etymological punning, and intertextual allusion deployed in the form of
riddles. As a locus of facts, data, and information, knowing in Muldoon’s poetry is contingent on
the play and ply of both the locally synchronic and the intertextually diachronic aspects of the
language used to structure it. These riddling dynamics are indefinitely played out in Muldoon’s
work, where reference and ambiguity as competing linguistic forces together constitute an
interminable weaving and unweaving of epistemological multiplicities. In Hill’s work, knowing is
disclosed negatively through a variety of apophatic tropes: combined with an aesthetics of
theological sublimity as well as the ethical demand for responsible language, Hill’s poetry
expresses knowing as an apophatic epistemological mystery.
Poems accomplish this interdisciplinary thinking about knowing through their resistance
to the rhetorics of representation, thematization, and closure, all of which are central features of
epistemological discourses that work to reveal, establish, or reinforce truth claims. In Zwicky,
Muldoon, and Hill, poetry complicates, problematizes, and resists the ontological simplifications
implied by the language-as-knowledge-producing model of epistemological discourse. By
exploring the paradigmatically gestalt, riddling, and apophatic qualities of poetry, this
dissertation provides insight into the contingencies of linguistically-derived truth, offering a view
of poetry not as an expression of knowledge but as “knowing language”