Field-wise wind estimation (also known as model-based wind estimation) is a sophisticated technique to derive wind estimates from radar backscatter measurements. In contrast to the more traditional method known as point-wise wind retrieval, field-wise techniques estimate wind field model parameters. In this way, neighboring wind vectors are jointly estimated, ensuring consistency. This work presents and implementation for field-wise wind retrieval for the SeaWinds scatterometer on the QuikSCAT satellite. Due to its sophistication, field-wise wind retrieval adds computational complexity and intensity. The tradeoffs necessary for practical implementations are examined and quantified. The Levenberg-Marquardt algorithm for minimizing the field-wise objective function is presented. As the objective function has several near-global local minima, several wind fields represent ambiguous wind field estimates. A deterministic method is proposed to ensure sufficient ambiguities are obtained. An improved method for selecting between ambiguous wind field estimates is also proposed. With a large set of Sea-Winds measurements and estimates available, the σ° measurement statistics are examined. The traditional noise model is evaluated for accuracy. A data-driven parameterization is proposed and shown to effectively estimate measurement bias and variance. The parameterized measurement model is used to generate Cramer-Rao bounds on estimator performance. Using the Cramer-Rao bound, field-wise and point-wise performances are compared.

Maximum likelihood symbol timing estimation for communication over a frequency non-selective MIMO fading channel is developed. The cases of known data (data-aided estimation) and unknown data (non-data-aided estimation) together with known channel and unknown channel are considered. The analysis shows that the log-likelihood functions and their approximations can be interpreted as SISO log-likelihood functions operating on each of the receive antennas. Previously published symbol timing estimators are shown to be special cases of the more general framework presented. Architectures based on both block processing and sequential processing using a discrete-time phase-locked loop are summarized. Performance examples over a MIMO channel based on measured data and on a simple stochastic MIMO channel model are given. These examples show that the mean-squared error performance of these techniques is not strongly dependent on the MIMO channel and is able to reach the Cramer Rao bound when sufficient complexity is applied.

Multi-Task Learning (MTL) has been an active research area in machine learning for two decades. By training multiple relevant tasks simultaneously with information shared across tasks, it is possible to improve the generalization performance of each task, compared to training each individual task independently. During the past decade, most MTL research has been based on the Regularization-Loss framework due to its flexibility in specifying various types of information sharing strategies, the opportunity it offers to yield a kernel-based methods and its capability in promoting sparse feature representations. However, certain limitations exist in both theoretical and practical aspects of Regularization-Loss-based MTL. Theoretically, previous research on generalization bounds in connection to MTL Hypothesis Space (HS)s, where data of all tasks are pre-processed by a (partially) common operator, has been limited in two aspects: First, all previous works assumed linearity of the operator, therefore completely excluding kernel-based MTL HSs, for which the operator is potentially non-linear. Secondly, all previous works, rather unnecessarily, assumed that all the task weights to be constrained within norm-balls, whose radii are equal. The requirement of equal radii leads to significant inflexibility of the relevant HSs, which may cause the generalization performance of the corresponding MTL models to deteriorate. Practically, various algorithms have been developed for kernel-based MTL models, due to different characteristics of the formulations. Most of these algorithms are a burden to develop and end up being quite sophisticated, so that practitioners may face a hard task in interpreting and implementing them, especially when multiple models are involved. This is even more so, when Multi-Task Multiple Kernel Learning (MT-MKL) models are considered. This research largely resolves the above limitations. Theoretically, a pair of new kernel-based HSs are proposed: one for single-kernel MTL, and another one for MT-MKL. Unlike previous works, we allow each task weight to be constrained within a norm-ball, whose radius is learned during training. By deriving and analyzing the generalization bounds of these two HSs, we show that, indeed, such a flexibility leads to much tighter generalization bounds, which often results to significantly better generalization performance. Based on this observation, a pair of new models is developed, one for each case: single-kernel MTL, and another one for MT-MKL. From a practical perspective, we propose a general MT-MKL framework that covers most of the prominent MT-MKL approaches, including our new MT-MKL formulation. Then, a general purpose algorithm is developed to solve the framework, which can also be employed for training all other models subsumed by this framework. A series of experiments is conducted to assess the merits of the proposed mode when trained by the new algorithm. Certain properties of our HSs and formulations are demonstrated, and the advantage of our model in terms of classification accuracy is shown via these experiments.

Outward Bound and the National Outdoor Leadership School (NOLS) are two of the United State’s largest providers of outdoor education and adventure recreation programs. While many studies have examined the outcomes of the different organizations individually, there is very little comparative research. This study compared the attributes, consequences, and values obtained by 510 participants of courses 14 days or longer in the Rocky Mountain region during the summer of 2006. Means-end theory was used to analyze data obtained from participants through personal interviews. Means-end theory links the physical objects or services, the means, with the outcomes and the personal values of the individual, the ends (Klenosky, Frauman, Norman, & Gengler 1998). The theory focuses on the interrelationship among attributes, consequences, and values, as three levels of abstraction (Goldenberg, Klenosky, O’Leary, & Templin, 2000). Data were collected using a convenience sampling method from OB schools in Leadville, Marble, and Silverton, Colorado and the NOLS headquarters in Lander, Wyoming, for a total of 162 OB participants and 348 NOLS participants. Interview questions were entered into LadderMap software, a program used to analyze means-end data. Content codes were developed and then tested by an additional researcher to measure intercoder reliability. An implication matrix was then created to tabulate the frequency of concepts being associated with one another. Hierarchical value maps (HVMs) were then created to graphically depict the themes and relationships that surfaced in the implication matrix. Seven HVMs were developed to visually present the data for all participants, all NOLS participants, all OB participants, OB males, OB females, NOLS males, and NOLS females. Though all HVMs were varied, some themes emerged by organization. For example, all NOLS participants had slightly more significant links to hard skills development than did their OB counterparts. NOLS and OB females stated being challenged and interactions as their most significant consequences yet corresponding males had slightly less emphasis on being challenged and more emphasis on new experience. This study produced more similarities than differences among the various subsets of the population. For example all HVMs showed a clear link from multiple attributes to independence, and ultimately to transference and additional values. The HVMs showed that new experiences, being challenged, and group interactions were significant components for all participants. The most common values obtained also demonstrate great similarity among participant demographics. These values include transference, sense of accomplishment, self respect/esteem/confidence, and self-awareness.

Protein-protein interactions build large networks which are essential in understanding complex diseases. Due to limitations of experimental methodology there are problems with large amounts of false negative and positive interactions; and a large gap in the amount of known interactions and structurally determined interactions. By using computational methods these problems can be alleviated. In this thesis the quality of a newly developed pipeline (InterPred) were investigated for its ability to generate coarse interaction models and score them. This ability was investigated by performing docking experiments in Rosetta on models generated in InterPred. The results suggest that InterPred is highly successful in generating good starting points for docking proteins in silico and to distinguish the quality of models.

Mémoire numérisé par la Direction des bibliothèques de l'Université de Montréal.

A numerical technique based on the theory of plasticity is developed to predict an optimum failure surface generatrix and concrete breakout capacity of single anchors away from edges under tensile loads. Concrete is regarded as a rigid, perfectly plastic material obeying a modified coulomb failure criteria with effective compressive and tensile strengths. The failure mode is idealized as an assemblage of two rigid blocks separated by failure surfaces of displacement discontinuity. Minimization of the collapse load predicted by the energy equation produces the optimum shape of the failure surface generatrix. A simplified solution is also developed by approximating the failure surface as two straight lines. The effect of different parameters on the concrete breakout capacity of anchors is reviewed using the developed mechanism analysis, ACI 318-05, and test results of 501 cast-in-place and 442 post-installed anchor specimens. The shape of failure surface and concrete breakout capacity of anchors predicted by the mechanism analysis are significantly affected by the ratio between effective tensile and compressive strengths of concrete. For anchors installed in concrete having a low ratio between effective tensile and compressive strengths, a much larger horizontal extent of failure planes in concrete surface is predicted by the mechanism analysis than recommended by ACI 318-05, similar to test results. Experimental concrete breakout capacity of anchors is closer to the prediction obtained from the mechanism analysis than ACI 318-05. ACI 318-05 provisions for anchors sharply underestimate the breakout capacity of cast-in-place and post-installed anchors having effective embedment depths exceeding 200 and 80 mm (7.87 to 3.15 in.), respectively, installed in concrete of compressive strength larger than 50 MPa (7250 psi).

CPU cache hierarchies are the central solution in bridging the memory wall. A proper understanding of how to trade-off their high cost against performance can lead to cost-savings without sacrificing performance.Due to the combinatorial nature of the problem, there exist a large number of configurations to investigate, making design space exploration slow and cumbersome. To improve this process, this Thesis develops and evaluates a model for optimally trading-off cost and performance of CPU cache hierarchies, named the Optimal Cache Problem (OCP), in the form of a Non-linear Integer Problem. A second goal of this work is the development of an efficient solver for the OCP, which was found to be a branch & bound algorithm and proven to function correctly. Experiments were conducted to empirically analyse and validate the model and to showcase possible use-cases. There, it was possible to ascribe the model outputs on measurable performance metrics. The model succeeded in formalising the inherent trade-off between cost and performance in a way that allows for an efficient and complete search of the configuration space of possible cache hierarchies. In future work, the model needs to be refined and extended to allow for the simultaneous analysis of multiple programs.

Private data analysis on statistical data has been addressed by many recent literatures. The goal of such analysis is to measure statistical properties of a database without revealing information of individuals who participate in the database. Differential privacy is a rigorous privacy definition that protects individual information using output perturbation: a differentially private algorithm produces statistically indistinguishable outputs no matter whether the database contains a tuple corresponding to an individual or not. It is straightforward to construct differentially private algorithms for many common tasks and there are published algorithms to support various tasks under differential privacy. However methods to design error-optimal algorithms for most non-trivial tasks are still unknown. In particular, we are interested in error-optimal algorithms for sets of linear queries. A linear query is a sum of counts of tuples that satisfy a certain condition, which covers the scope of many aggregation tasks including count, sum and histogram. We present the matrix mechanism, a novel mechanism for answering sets of linear queries under differential privacy. The matrix mechanism makes a clear distinction between a set of queries submitted by users, called the query workload, and an alternative set of queries to be answered under differential privacy, called the query strategy. The answer to the query workload can then be computed using the answer to the query strategy. Given a query workload, the query strategy determines the distribution of the output noise and the power of the matrix mechanism comes from adaptively choosing a query strategy that minimizes the output noise. Our analyses also provide a theoretical measure to the quality of different strategies for a given workload. This measure is then used in accurate and approximate formulations to the optimization problem that outputs the error-optimal strategy. We present a lower bound of error to answer each workload under the matrix mechanism. The bound reveals that the hardness of a query workload is related to the spectral properties of the workload when it is represented in matrix form. In addition, we design an approximate algorithm, which generates strategies generated by our a out perform state-of-art mechanisms over (epsilon, delta)-differential privacy. Those strategies lead to more accurate data analysis while preserving a rigorous privacy guarantee. Moreover, we also combine the matrix mechanism with a novel data-dependent algorithm, which achieves differential privacy by adding noise that is adapted to the input data and to the given query workload.

<p> A solution to an industrial problem of designing a hydrostatic drive system for logging vehicles is presented. A computer program using Land and Doig's method of Branch and Bound Mixed Integer Programming is used to obtain an optimal solution to the problem. A comprehensive users guide to allow the use of the program by unsophisticated users is provided. </p> <p> An attempted alternate method of solution using the Gomory method of Integer Programming is presented and reasons for its failure discussed. </p> / Thesis / Master of Engineering (MEngr)