The thesis contains four chapters on the structure and predictability of individual differences Chapter 1. Re-analyses data from Holt and Laury's (2002) risk aversion experiments. Shows that big-stakes hypothetical payoffs are better than small-stakes real-money payoffs for predicting choices in big-stakes real-money gambles (in spite of the presence of hypothetical bias). Argues that hypothetical bias is a problem for calibration of mean preferences but not for prediction of the rank order of subjects' preferences. Chapter 2. Describes an experiment: Participants were given personality tests and played a series of dictator and response games over a two week period. It was found that social preferences are one-dimensional, stable across a two-week interval and significantly related to the Big Five personality traits. Suggestions are given about ways to modify existing theories of social preference to accommodate these findings. Chapter 3. Applies a novel statistical technique (spectral clustering) to a personality data set for the first time. Finds the HEXACO six-factor structure in an English-language five-factor questionnaire for the first time. Argues that the emphasis placed on weak relationships is critical to settling the dimensionality debate within personality theory, and that spectral clustering provides a more useful perspective on personality data than does traditional factor analysis. Chapter 4. Outlines the relevance of extraversion for economics, and sets up a model to argue that personality differences in extraversion may have evolved through something akin to a war of attrition. This model implies a positive relationship between extraversion and risk aversion, and a U-shaped relationship between extraversion and loss aversion.

This paper explores the potential applications of existing spectral clustering algorithms to real life problems through experiments on existing road traffic data. The analysis begins with an overview of previous unsupervised machine learning techniques and constructs an effective spectral clustering algorithm that demonstrates the analytical power of the method. The paper focuses on the spectral embedding method’s ability to project non-linearly separable, high dimensional data into a more manageable space that allows for accurate clustering. The key step in this method involves solving a normalized eigenvector problem in order to construct an optimal representation of the original data. While this step greatly enhances our ability to analyze the relationships between data points and identify the natural clusters within the original dataset, it is difficult to comprehend the eigenvalue representation of the data in terms of the original input variables. The later sections of this paper will explore how the careful framing of questions with respect to available data can help researchers extract tangible decision driving results from real world data through spectral clustering analysis.

We have previously introduced a high throughput multiplexing computational spectral imaging device. The device measures scalar projections of pseudo-arbitrary spectral filters at each spatial pixel. This paper discusses simulation and initial experimental progress in performing computational spectral unmixing by taking advantage of the natural sparsity commonly found in the fractional abundances. The simulation results show a lower unmixing error compared to traditional spectral imaging devices. Initial experimental results demonstrate the ability to directly perform spectral unmixing with less error than multiplexing alone.

As the availability of unstructured data on the web continues to increase, it is becoming increasingly necessary to develop machine learning methods that rely less on human annotated training data. In this thesis, we present methods for learning from weakly labeled data. We present a unifying framework to understand weakly labeled data in terms of bias and noise and identify methods that are well suited to learning from certain types of weak labels. To compensate for the tremendous sizes of weakly labeled datasets, we leverage computationally efficient and statistically consistent spectral methods. Using these methods, we present results from four diverse, real-world applications coupled with a unifying simulation environment. This allows us to make general observations that would not be apparent when examining any one application on its own. These contributions allow us to significantly improve prediction when labeled data is available, and they also make learning tractable when the cost of acquiring annotated data is prohibitively high.

Probabilistic modeling with latent variables is a powerful paradigm that has led to key advances in many applications such natural language processing, text mining, and computational biology. Unfortunately, while introducing latent variables substantially increases representation power, learning and modeling can become considerably more complicated. Most existing solutions largely ignore non-identifiability issues in modeling and formulate learning as a nonconvex optimization problem, where convergence to the optimal solution is not guaranteed due to local minima. In this thesis, we propose to tackle these problems through the lens of linear/multi-linear algebra. Viewing latent variable models from this perspective allows us to approach key problems such as structure learning and parameter learning using tools such as matrix/tensor decompositions, inversion, and additive metrics. These new tools enable us to develop novel solutions to learning in latent variable models with theoretical and practical advantages. For example, our spectral parameter learning methods for latent trees and junction trees are provably consistent, local-optima-free, and 1-2 orders of magnitude faster thanEMfor large sample sizes. In addition, we focus on applications in Natural Language Processing, using our insights to not only devise new algorithms, but also to propose new models. Our method for unsupervised parsing is the first algorithm that has both theoretical guarantees and is also practical, performing favorably to theCCMmethod of Klein and Manning. We also developed power low rank ensembles, a framework for language modeling that generalizes existing n-gram techniques to non-integer n. It consistently outperforms state-of-the-art Kneser Ney baselines and can train on billion-word datasets in a few hours.

The present work focuses on two strategies contributing to the development of high efficiency, cost-effective photovoltaic (PV) technology for renewable energy generation: the design of new materials offering enhanced opto-electronic performance and the investigation of material degradation processes and their role in predicting the long-term reliability of PV modules in the field. The first portion of the present work investigates the integration of a novel CdTe-ZnO nanocomposite material as a spectral sensitizer component within a thin-film, hybrid heterojunction (HJ) PV device structure. Quantum-scale semiconductors have the potential to improve PV device performance through enhanced spectral absorption and photocarrier transport. This is realized via appropriate design of the semiconductor nanophase (providing tunable spectral absorption) and its spatial distribution within an electrically active matrix (providing long-range charge transport). Here, CdTe nanocrystals, embedded in an electrically active ZnO matrix, form a nanocomposite (NC) offering control of both spectral absorption and photocarrier transport behavior through the manipulation of nanophase assembly (ensemble effects). A sequential radio- frequency (RF) magnetron sputter deposition technique affords the control of semiconductor nanophase spatial distribution relative to the HJ plane in a hybrid, ZnO-P3HT test structure. Energy conversion performance (current density-voltage (J-V) and external quantum efficiency (EQE) response) was examined as a function of the location of the CdTe nanophase absorber region using both one dimensional solar cell capacitance simulator (SCAPS) and the experimental examination of analogous P3HT-ZnO based hybrid thin films. Enhancement in simulated EQE over a spectral range consistent with the absorption region of the CdTe nanophase (i.e. 400–475 nm) is confirmed in the experimental studies. Moreover, a trend of decreasing quantum efficiency in this spectral range with increasing separation between the CdTe nanophase region and the heterojunction plane is observed. The results are interpreted in terms of carrier scattering/recombination length mitigating the successful transport of carriers across the junction. The second portion of the research addresses the need for robust PV performance in commercial module as a primary contributor to cost-effective operation in both distributed systems and utility scale generation systems. The understanding of physical and chemical mechanisms resulting in the degradation of materials of construction used in PV modules is needed to understand the contribution of these processes to module integrity and performance loss with time under varied application environments. In this context, the second part of present study addresses microcracking in Si–an established degradation process contributing to PV module power loss. The study isolates microcrack propagation in single-crystal Si, and investigates the effect of local environment (temperature, humidity) on microcrack elongation under applied strains. An investigation of microindenter-induced crack evolution with independent variation of both temperature and vapor density was pursued in PV-grade Si wafers. Under static tensile strain conditions, an increase in sub-critical crack elongation with increasing atmospheric water content was observed. To provide further insight into the potential physical and chemical conditions at the microcrack tip, micro-Raman measurements were performed. Preliminary results confirm a spatial variation in the frequency of the primary Si vibrational resonance within the crack-tip region, associated with local stress state, whose magnitude is influenced by environmental conditions during the period of applied static strain. The experimental effort was paired with molecular dynamics (MD) investigations of microcrack evolution in single-crystal Si to furnish additional insight into mechanical contributions to crack elongation. The MD results demonstrate that crack-tip energetics and associated cracking crystal planes and morphology are intimately related to the crack and applied strain orientations with respect to the principal crystallographic axes. The resulting fracture surface energy and the stress-strain response of the Si under these conditions form the basis for preliminary micro-scale peridynamics (PD) simulations of microcrack development under constant applied strain. These efforts were integrated with the experimental results to further inform the mechanisms contributing to this important degradation mode in Si-based photovoltaics.

Southern Louisiana is experiencing a dramatic loss of freshwater wetlands as a result of natural and man-made changes in the landscape. Multitempral remotely sensed data were used to examine the impact of the Caernarvon Freshwater Diversion Structure, built in 1991 to divert water to Breton Sound. Satellite imagery data covering the period from 1974 to 2006 were analyzed by computing several spectral indices including NDVI, VI, IR/R, Sqrt IR/R, T-NDVI, and NDWI, as well as principle component analysis. The resulting enhanced images were classified into two classes, vegetation or open water. The ratios of vegetation to open water were then calculated and the changes graphed over the 1974-2006 timeframe. The results indicated that despite the infusion of freshwater, the open water portion of the Breton Sound area continued to expand, indeed the expansion rate increased from approximately 0.25% per year before construction of Caernarvon to 0.45% per year after construction.

Industrial machines during their operation generate vibration due to dynamic forces acting on the machines. This vibration may create noise, abrasion in the machine parts, mechanical fatigue, degrade performance, transfer to other machines via floor or walls and may cause complete shutdown of the machine. To limit the vibration pre-installation, vibration isolation measures are usually employed in workshops and industrial units. However, such vibration isolation may not be sufficient due to varying operating and physical conditions, such as machine ageing, structural changes and new installations etc. Therefore, it is important to assess the quantity of vibration generated and transmitted during true operating conditions. The thesis work is aimed at the estimation of vibrational transmissibility or transfer from industrial machines to floor and to other adjacent installed machines. This study of transmissibility is based on the measurement and analysis of various spectral estimation tools such as Power Spectral Density (PSD), Frequency Response Function (FRF) and Coherence Function. The overall study is divided into three major steps. Firstly, the initial measurements are carried in BTH on simple Single Degree of Freedom (SDOF) systems to gain confidence in measurement and analysis. Then the measurements are performed on a Lathe machine “Quick Turn Nexus 300-II” in a laboratory at BTH. Finally, the measurements are taken on the machines of an Industrial workshop (KOSAB). The analysis results revealed that vibration measurements in industry are challenging and not easy as measurement in labs. Measurements are contaminated by noise from other machines, which degrade the coherence function. However, vibration transferred from one machine to the floor or other machines may be studied using FRF and PSD. Appropriate further isolations may be employed based on the spectral analysis.

To accommodate future power demands, wave energy converters will be deployed in arrays, but largely unanswered questions of the annual energy production and environmental impact of such installations present regulatory dilemmas. In recent years, Sandia National Laboratories (SNL) has developed a modified version of the Simulating Waves Nearshore (SWAN) wave model to simulate WEC energy extraction in a propagating wave field. This thesis presents a novel WEC meta-model that calculates the power intercepted by a WEC from the incident wave field. Two representations were developed with which a user could model a WEC’s impact on the incident waves in a spectral wave model. These alterations are based on power a WEC captures from the sea and power dissipated by hydrodynamic losses calculated in an external six degree of freedom (DOF) time domain WEC simulation. The two WEC meta-models were compared in terms of significant wave height reduction in the WEC’s lee and annual power production. The first WEC representation removes a constant percentage of power from each frequency bin while the second representation employs frequency dependent energy extraction. The representations were then applied in modelling a 54 MW WEC array off of Amphitrite Bank on the West Coast of Vancouver Island. Over the course of a year, the power captured by a farm when represented with a constant percentage extraction is reduced by 2.9% while a frequency dependent percentage extraction reduced the farm’s total captured power by 2.3% when compared to the reference case. Similarly small changes were observed in significant wave height reductions. The significant wave height in the lee of a farm was reduced by less than 2% for both representations at the shoreline, approximately six kilometres behind the farm. / Graduate / 0775, 0547, 0548 / eluczko91@gmail.com

Thesis (Ph.D.)--University of the Witwatersrand, Faculty of Science, School of Mathematics, 2017. / ForwardandinversespectralproblemsconcerningSturm-Liouvilleoperatorswithoutdiscontinuitieshavebeenstudiedextensively. Bycomparison,therehasbeenlimitedworktacklingthecase where the eigenfunctions have discontinuities at interior points, a case which appears naturally in physical applications. We refer to such discontinuity conditions as transmission conditions. We consider Sturm-Liouville problems with transmission conditions rationally dependent on the spectral parameter. We show that our problem admits geometrically double eigenvalues, necessitating a new analysis. We develop the forward theory associated with this problem and also consider a related inverse problem. In particular, we prove a uniqueness result analogous to that of H. Hochstadt on the determination of the potential from two sequences of eigenvalues. In addition, we consider the problem of extending Sturm’s oscillation theorem, regarding the number of zeroes of eigenfunctions, from the classical setting to discontinuous problems with general constant coefficient transmission conditions / GR2018