As the current rate of improvement in processor performance far exceeds the rate of memory performance, memory latency is the dominant overhead in many performance critical applications. In many cases, automatic compiler-based approaches to improving memory performance are limited and programmers frequently resort to manual optimisation techniques. However, this process is tedious and time-consuming. Furthermore, a diverse range of a rapidly evolving hardware makes the optimisation process even more complex. It is often hard to predict the potential benefits from different optimisations and there are no simple criteria to stop optimisations i.e. when optimal memory performance has been achieved or sufficiently approached. This thesis presents a platform independent optimisation approach for numerical applications based on iterative feedback-directed program restructuring using a new reasonably fast and accurate performance prediction technique for guiding optimisations. New strategies for searching the optimisation space, by means of profiling to find the best possible program variant, have been developed. These strategies have been evaluated using a range of kernels and programs on different platforms and operating systems. A significant performance improvement has been achieved using new approaches when compared to the state-of-the-art native static and platform-specific feedback directed compilers.

In this thesis, a new smoothing penalty algorithm is introduced to solve a mathematical program with equilibrium constraints (MPEC). By smoothing the exact penalty function, an MPEC is reformulated as a series of subprograms which belong to a class of MPECs with simple linear complementarity constraints. To deal with the subproblems, a hybrid algorithm is proposed, which combines the active set algorithm, the 6-active search algorithm and the PSQP algorithm. It is shown that the smoothing penalty algorithm converges globally to a M-stationary point of MPEC under weak conditions. Supervisor: Dr. Jane Ye (Department of Mathematics and Statistics) Co-Supervisor: Dr. Wu-Sheng Lu (Department of Electrical and Computer Engineering)

Modeling sediment erosion is important in a wide range of environmental problems. The effects of various environmental factors on erosion rates have been studied, but the effects of surface slope on erosion rates of a wide range of sediments have not been quantified. The effects of surface slope, both in the direction of flow (pitch) and perpendicular to the flow (roll), on erosion rates of quartz particles were investigated using the Sediment Erosion at Depth Flume (Sedflume). / US Navy (USN) author.

In this thesis, a wavelet-based prediction method is developed for concealing packet-loss effects in wireless channels. The proposed method utilizes a wavelet decomposition algorithm in order to process the data and then applies the well known linear prediction technique to estimate one or more approximation coefficients as necessary at the lowest resolution level. The predicted sample stream is produced by using the predicted approximation coefficients and by exploiting certain sample value patterns in the detail coefficients. In order to test the effectiveness of the proposed scheme, a wireless channel based on a three-state Markov model is developed and simulated. Simulation results for transmission of image and speech packet streams over a wireless channel are reported for both the wavelet-based prediction and direct linear prediction. In all the simulations run in this work, the wavelet-based method outperformed the direct linear prediction method. / Hellenic Navy author.

This paper explores ways in which civil conflict can be simulated using numerical methods. A general two-party model of conflict is developed by extending an approach proposed by [Christia, F., (2012), Alliance Formation in Civil Wars, Cambridge University Press, New York], which is based on a metric of the 'relative power' that exists between the state and a rebel group. Various definitions of relative power are considered and one of these is chosen to illustrate different types of two-sided armed conflict, namely direct-fire, guerrilla and asymmetric warfare. The additional suggestion of Christia that random or stochastic events can lead to unexpected conflict outcomes is also further extended in this paper. The inclusion in the model of terms describing concurrent rebel recruitment of civilians and state deployment of troops are then described. Examples are presented for various hypothetical cases. It is demonstrated that numerical simulation techniques have great potential for modelling civil war. The Christia approach is shown to provide an excellent basis from which numerical models of civil conflict can be built and from which the progress of a conflict can usefully be visualised graphically.

Self-Organizing Maps (SOM) can provide a visualization for multi-dimensional data with two dimensional mappings. By applying unsupervised learning techniques to SOM representations, we can further enhance visual inspection for change detection. In order to obtain a more accurate measurement for the changes of self-organizing maps beyond simple visual inspection, we introduce the Gaussian Mixture Model (GMM) and Kullback-Leibler Divergence (KLD) on top of SOM trained maps. The main contribution in this dissertation focuses on adding numerical methods to SOM algorithms, with anomaly detection as example domain. Through extensive traced-based simulations, it is observed that our techniques can uncover anomalies with an accuracy of 100% at an anomaly mixture-rate as low as 12% from the CTU-13 dataset. Tuning of the KLD threshold further reduces the mixture-rate to 7%, significantly augmenting visual inspection to assist in detecting low-rate anomalies. Suitable hierarchical and distributed SOM-based approaches are also explored, along with other approaches in the literature. Hierarchies in SOM can show the correlations among the neural cells on the self-organizing maps. In order to obtain a higher accuracy for anomaly detection, a new dimension of labels is suggested to be added in the second layer of SOM training. Also for more general distributed SOM-based algorithms, we investigate the use of principal component analysis (PCA) for the separation of dimensions. With the transformed dataset from PCA, the inner dependencies can be reserved in a manageable scale. As a case study, this dissertation uses a SOM-based approach for anomaly detection in Supervisory Control And Data Acquisition (SCADA) networks. We further investigate the use of SOM for the Quality of Service (QoS) in the scenario of wireless SCADA networks. Solving the problem of long computing time of optimizing the cached contents, the new SOM-based approach can also learn and predict the sub-optimal locations for the caching while maintaining a prediction error of 28%. / Graduate

The main objectives of this study were to obtain new experimental data for conditions not previously tested for discharging two-phase flow through two 6.35 mm diameter branches with centrelines falling in an inclined plane and to assess the applicability of ANSYS CFX in modelling discharging two-phase flow through various single and multiple branch geometries. The present results are relevant to many industrial applications including headers and manifolds, multichannel heat exchangers and small breaks in horizontal pipes. In the experimental investigation, onsets of liquid and gas entrainment data were obtained, analyzed and correlated for two different branch spacings and two different angles between the branches. For each combination of branch spacing and angle between the branches, a wide range of Froude numbers was used. Two-phase mass flow rate and quality results were also obtained and analyzed for a range of interface heights for 16 different combinations of branch spacing, inclination angle, test section pressure and pressure drop across each branch. New correlations were developed to predict the dimensionless mass flow rate and the quality. The new correlations show good agreement with the present data and with previous correlations. Using ANSYS CFX, the inhomogeneous, free surface model was used to model discharging two-phase flow through horizontal branches with the following configurations: a single short branch of square cross-section (G1), a single long branch of square cross-section (G2), a single long branch of round cross section (G3), two round branches located one on top of the other (G4), and two round branches with their centreplane angled 30° from the horizontal. For these five geometries, results were compared with previous and present results and showed good agreement for Geometries G1, G2, G3 and G5. For Geometry G4, CFX was unable to predict results when the interface was located such that the two phases flowed through both branches simultaneously. Under these conditions, the flow phenomena were too complicated and further investigations are necessary. / February 2017

This is a study of numerical control, its application, and the design of a numerical control program for a drilling machine. The study sought to obtain information on the function of numerical control; the different types of numerical control used by industry; and, the advantages of using numerical control.

Thesis advisor: Sara Cordes / Human infants show relatively sophisticated abilities to track and use proportional information. However, by the age of 6, children tend to make predictable errors in their proportional reasoning and later encounter significant challenges in many aspects of formal fraction learning. Thus, one of the central questions motivating this research is to identify the factors leading to these difficulties, in light of evidence of early intuitions about these concepts. In the current dissertation, I address this question by investigating the tradeoff between attending to proportional magnitude information and discrete numerical information about the components (termed “numerical interference”) across both spatial (i.e., area models, number lines) and symbolic (fractions, decimals) representations of proportion information. These explorations focus on young children (5-7 year olds) who have yet to receive formal fraction instruction, older children (9-12 year olds) who are in the process of learning these concepts, and adults who have already learned formal fractions. In Project 1, I investigated how older children and adults map between symbolic and spatial representations, particularly focusing on their strategies in highlighting componential information versus magnitude information when solving these mapping tasks. In Projects 2 and 3, I explore the malleability of individual differences in this numerical interference in 4- to 7-year-old children. Across the three projects, I suggest that although numerical interference does impact proportional reasoning, this over-attention to number can be reduced through modifying early experiences with proportional information. These findings have implications for education and the way we conceptualize numerical interference more generally. / Thesis (PhD) — Boston College, 2017. / Submitted to: Boston College. Graduate School of Arts and Sciences. / Discipline: Psychology.

A dissertation submitted to the Faculty of Science, University of the Witwatersrand, in fulfillment of the requirements for the degree of Master of Science, May 14, 2013. / In this dissertation, we investigate the parabolic mixed derivative diffusion equation modeling the viscous and viscoelastic effects in a non-Newtonian viscoelastic fluid. The model is analytically considered using Fourier and Laplace transformations. The main focus of the dissertation, however, is the implementation of the Peaceman-Rachford Alternating Direction Implicit method. The one-dimensional parabolic mixed derivative diffusion equation is extended to a two-dimensional analog. In order to do this, the two-dimensional analog is solved using a Crank-Nicholson method and implemented according to the Peaceman- Rachford ADI method. The behaviour of the solution of the viscoelastic fluid model is analysed by investigating the effects of inertia and diffusion as well as the viscous behaviour, subject to the viscosity and viscoelasticity parameters. The two-dimensional parabolic diffusion equation is then implemented with a high-order method to unveil more accurate solutions. An error analysis is executed to show the accuracy differences between the numerical solutions of the general ADI and high-order compact methods. Each of the methods implemented in this dissertation are investigated via the von-Neumann stability analysis to prove stability under certain conditions.