Gaussian processes:iterative sparse approximations

In recent years there has been an increased interest in applying non-parametric methods to real-world problems. Significant research has been devoted to Gaussian processes (GPs) due to their increased flexibility when compared with parametric models. These methods use Bayesian learning, which generally leads to analytically intractable posteriors. This thesis proposes a two-step solution to construct a probabilistic approximation to the posterior. In the first step we adapt the Bayesian online learning to GPs: the final approximation to the posterior is the result of propagating the first and second moments of intermediate posteriors obtained by combining a new example with the previous approximation. The propagation of em functional forms is solved by showing the existence of a parametrisation to posterior moments that uses combinations of the kernel function at the training points, transforming the Bayesian online learning of functions into a parametric formulation. The drawback is the prohibitive quadratic scaling of the number of parameters with the size of the data, making the method inapplicable to large datasets. The second step solves the problem of the exploding parameter size and makes GPs applicable to arbitrarily large datasets. The approximation is based on a measure of distance between two GPs, the KL-divergence between GPs. This second approximation is with a constrained GP in which only a small subset of the whole training dataset is used to represent the GP. This subset is called the em Basis Vector, or BV set and the resulting GP is a sparse approximation to the true posterior. As this sparsity is based on the KL-minimisation, it is probabilistic and independent of the way the posterior approximation from the first step is obtained. We combine the sparse approximation with an extension to the Bayesian online algorithm that allows multiple iterations for each input and thus approximating a batch solution. The resulting sparse learning algorithm is a generic one: for different problems we only change the likelihood. The algorithm is applied to a variety of problems and we examine its performance both on more classical regression and classification tasks and to the data-assimilation and a simple density estimation problems.

Lobe connections and lobe crowding are associated with growth rate in the lichen Xanthoparmelia conspersa

The association between lobe connections and the degree of lobe crowding and radial growth was studied in thalli of the foliose lichen Xanthoparmelia conspersa. In 35 thalli, 15% of the lobes were not physically connected to either of their neighbours before the lobes merged into the centre of the thallus. Twenty-five percent of the lobes were connected in pairs and 29% in groups of three. Approximately 5% of the lobes were interconnected in larger groups of six or more. The mean number of lobes per group in a thallus was positively correlated with thallus diameter and with the degree of lobe growth variation but was unrelated to annual radial growth rate (RGR). The degree of crowding of the lobes in a thallus was defined as a 'crowding index', viz., the product of lobe density and mean lobe width. Crowding index increased rapidly with size in smaller thalli but changed less with size in larger thalli. Crowding index was positively correlated with RGR but was unrelated to lobe growth variation. Lobes removed from large thalli and glued in various configurations to simulate different degrees of crowding did not demonstrate an association between lobe crowding and RGR over one year. These results suggest that the pattern of lobe connectivity of a thallus is associated with lobe growth variation in X. conspersa. The degree of lobe crowding is associated with the increase in RGR with thallus size in smaller thalli and by restricting lobe width, could also be a factor associated with the more constant growth of larger thalli.

Florid prion protein (PrP) plaques in patients with variant Creutzfeldt-Jakob disease (vCJD) are spatially related to blood vessels

This study tested the hypothesis that variations in the density of the florid prion protein (PrP) plaques in the brain of patients with variant Creutzfeldt-Jakob disease (vCJD) were spatially related to blood vessels. In 81% of areas of the cerebral cortex sampled and in 37% of the remaining areas, which included the hippocampus, dentate gyrus, and cerebellum, there was a positive spatial correlation between the density of the florid plaques and the larger blood vessel profiles. The frequency of the positive spatial correlations was similar in different anatomical areas of the cerebral cortex and in the upper compared with the lower cortical laminae. The data support the hypothesis that the florid plaques cluster around the larger blood vessels in vCJD, the density of associated plaques increasing with vessel size. The development of florid plaques close to blood vessels may be due to factors associated with the blood vessels that enhance the aggregation of PrP to form the dense cores of florid plaques and is unlikely to reflect the haematogenous spread of PrP into the brain.

Modelling of induction motors for system faults and transient stability studies

The research carried out in this thesis was mainly concerned with the effects of large induction motors and their transient performance in power systems. Computer packages using the three phase co-ordinate frame of reference were developed to simulate the induction motor transient performance. A technique using matrix algebra was developed to allow extension of the three phase co-ordinate method to analyse asymmetrical and symmetrical faults on both sides of the three phase delta-star transformer which is usually required when connecting large induction motors to the supply system. System simulation, applying these two techniques, was used to study the transient stability of a power system. The response of a typical system, loaded with a group of large induction motors, two three-phase delta-star transformers, a synchronous generator and an infinite system was analysed. The computer software developed to study this system has the advantage that different types of fault at different locations can be studied by simple changes in input data. The research also involved investigating the possibility of using different integrating routines such as Runge-Kutta-Gill, RungeKutta-Fehlberg and the Predictor-Corrector methods. The investigation enables the reduction of computation time, which is necessary when solving the induction motor equations expressed in terms of the three phase variables. The outcome of this investigation was utilised in analysing an introductory model (containing only minimal control action) of an isolated system having a significant induction motor load compared to the size of the generator energising the system.

Low bit-rate speech encoding for digital mobile radio

The need for low bit-rate speech coding is the result of growing demand on the available radio bandwidth for mobile communications both for military purposes and for the public sector. To meet this growing demand it is required that the available bandwidth be utilized in the most economic way to accommodate more services. Two low bit-rate speech coders have been built and tested in this project. The two coders combine predictive coding with delta modulation, a property which enables them to achieve simultaneously the low bit-rate and good speech quality requirements. To enhance their efficiency, the predictor coefficients and the quantizer step size are updated periodically in each coder. This enables the coders to keep up with changes in the characteristics of the speech signal with time and with changes in the dynamic range of the speech waveform. However, the two coders differ in the method of updating their predictor coefficients. One updates the coefficients once every one hundred sampling periods and extracts the coefficients from input speech samples. This is known in this project as the Forward Adaptive Coder. Since the coefficients are extracted from input speech samples, these must be transmitted to the receiver to reconstruct the transmitted speech sample, thus adding to the transmission bit rate. The other updates its coefficients every sampling period, based on information of output data. This coder is known as the Backward Adaptive Coder. Results of subjective tests showed both coders to be reasonably robust to quantization noise. Both were graded quite good, with the Forward Adaptive performing slightly better, but with a slightly higher transmission bit rate for the same speech quality, than its Backward counterpart. The coders yielded acceptable speech quality of 9.6kbps for the Forward Adaptive and 8kbps for the Backward Adaptive.

Size matters, but not for everyone:Individual differences for contrast discrimination

It is very well known that contrast detection thresholds improve with the size of a grating-type stimulus, but it is thought that the benefit of size is abolished for contrast discriminations well above threshold (e.g., Legge, G. E., & Foley, J. M. (1980)]. Here we challenge the generality of this view. We performed contrast detection and contrast discrimination for circular patches of sine wave grating as a function of stimulus size. We confirm that sensitivity improves with approximately the fourth-root of stimulus area at detection threshold (a log-log slope of -0.25) but find individual differences (IDs) for the suprathreshold discrimination task. For several observers, performance was largely unaffected by area, but for others performance first improved (by as much as a log-log slope of -0.5) and then reached a plateau. We replicated these different results several times on the same observers. All of these results were described in the context of a recent gain control model of area summation [Meese, T. S. (2004)], extended to accommodate the multiple stimulus sizes used here. In this model, (i) excitation increased with the fourth-root of stimulus area for all observers, and (ii) IDs in the discrimination data were described by IDs in the relation between suppression and area. This means that empirical summation in the contrast discrimination task can be attributed to growth in suppression with stimulus size that does not keep pace with the growth in excitation. © 2005 ARVO.

Factors associated with degeneration of the thallus centre in foliose lichens

Degeneration of the older parts of foliose lichen thalli often lead to the formation of a space or 'window' in the centre of the colonies. The percentage of thalli of different size which exhibited 'windows' was studied in twenty saxicolous lichen populations in south Gwynedd, Wales. The proportion of thalli with 'windows' increased with thallus size. The size class at which 50% and 100% of thalli exhibited 'windows' varied between populations. Differences between populations were not correlated with distance from the sea, aspect, slope or porosity of the substrate or the total number of lichen species present. However, a higher percentage of smaller thalli had 'windows' on rock surfaces with a greater lichen cover. There were no significant differences in the levels of Ca, Mg, Cu or Zn in large (>4 cm) and small (<2 cm) Parmelia conspersa (Ehrh. ex Ach.) Ach. thalli or in the centres and marginal lobes of these thalli. The concentration of ribitol, arabitol and mannitol was significantly reduced in the centre of large thalli compared with the margin of large thalli and the centre of small thalli. However, carbohydrate levels were similar in the centre of large thalli and the margin of small thalli. The data suggest that loss of the thallus centre is a degenerative process related to thallus size. In the field, the formation of 'windows' may be related to the intensity of competition on a substrate. Central degeneration was not associated with a deficiency or an accumulation of Ca, Mg, Cu and Zn in the thallus centre. However, degeneration may be associated with a reduction in carbohydrates in the centre compared with the marginal lobes.

A study of cluster analysis techniques and their applications

This thesis seeks to describe the development of an inexpensive and efficient clustering technique for multivariate data analysis. The technique starts from a multivariate data matrix and ends with graphical representation of the data and pattern recognition discriminant function. The technique also results in distances frequency distribution that might be useful in detecting clustering in the data or for the estimation of parameters useful in the discrimination between the different populations in the data. The technique can also be used in feature selection. The technique is essentially for the discovery of data structure by revealing the component parts of the data. lhe thesis offers three distinct contributions for cluster analysis and pattern recognition techniques. The first contribution is the introduction of transformation function in the technique of nonlinear mapping. The second contribution is the us~ of distances frequency distribution instead of distances time-sequence in nonlinear mapping, The third contribution is the formulation of a new generalised and normalised error function together with its optimal step size formula for gradient method minimisation. The thesis consists of five chapters. The first chapter is the introduction. The second chapter describes multidimensional scaling as an origin of nonlinear mapping technique. The third chapter describes the first developing step in the technique of nonlinear mapping that is the introduction of "transformation function". The fourth chapter describes the second developing step of the nonlinear mapping technique. This is the use of distances frequency distribution instead of distances time-sequence. The chapter also includes the new generalised and normalised error function formulation. Finally, the fifth chapter, the conclusion, evaluates all developments and proposes a new program. for cluster analysis and pattern recognition by integrating all the new features.

A study of adaptive load balancing algorithms for distributed systems

With the advent of distributed computer systems with a largely transparent user interface, new questions have arisen regarding the management of such an environment by an operating system. One fertile area of research is that of load balancing, which attempts to improve system performance by redistributing the workload submitted to the system by the users. Early work in this field concentrated on static placement of computational objects to improve performance, given prior knowledge of process behaviour. More recently this has evolved into studying dynamic load balancing with process migration, thus allowing the system to adapt to varying loads. In this thesis, we describe a simulated system which facilitates experimentation with various load balancing algorithms. The system runs under UNIX and provides functions for user processes to communicate through software ports; processes reside on simulated homogeneous processors, connected by a user-specified topology, and a mechanism is included to allow migration of a process from one processor to another. We present the results of a study of adaptive load balancing algorithms, conducted using the aforementioned simulated system, under varying conditions; these results show the relative merits of different approaches to the load balancing problem, and we analyse the trade-offs between them. Following from this study, we present further novel modifications to suggested algorithms, and show their effects on system performance.

Characterization of the hairpin vortex solution in plane Couette flow

Quantitative evidence that establishes the existence of the hairpin vortex state (HVS) in plane Couette flow (PCF) is provided in this work. The evidence presented in this paper shows that the HVS can be obtained via homotopy from a flow with a simple geometrical configuration, namely, the laterally heated flow (LHF). Although the early stages of bifurcations of LHF have been previously investigated, our linear stability analysis reveals that the root in the LHF yields multiple branches via symmetry breaking. These branches connect to the PCF manifold as steady nonlinear amplitude solutions. Moreover, we show that the HVS has a direct bifurcation route to the Rayleigh-Bénard convection. © 2010 The American Physical Society.

Novel bioadhesive formulations for mucosal and parenteral delivery of vaccines

In this work we have established the efficient mucosal delivery of vaccines using absorption enhancers and chitosan. In addition, the use of chitosan was shown to enhance the action of other known adjuvants, such as CTB or Quil-A. Collectively, the results presented herein indicate that chitosan has excellent potential as a mucosal adjuvant. We have evaluated a number of absorption enhancers for their adjuvant activity in vivo. Polyornithine was shown to engender high scrum immune reasons to nasally delivered antigens, with higher molecular weight polyornithine facilitating the best results. We have demonstrated for the first time that vitamin E TPGS can act as mucosal adjuvant. Deoxycholic acid, cyclodextrins and acylcarnitines were also identified as effective mucosal adjuvants and showed enhanced immune responses to nasally delivered TT, DT and Yersinia pestis V and F1 antigens. Previously, none of these agents, common in their action as absorption enhancing agents, have been shown to have immunopotentiating activity for mucosal immunisation. We have successfully developed novel surface modified microspheres using chitosan as an emulsion stabiliser during the preparation of PLA microspheres. It was found that immune responses could be substantially increased, effectively exploiting the immunopenetrating characteristics of both chitosan and PLA microspheres in the same delivery vehicle. In the same study, comparison of intranasal and intramuscular routes of administration showed that with these formulations, the nasal route could be as effective as intramuscular delivery, highlighting the potential of mucosal administration for these particulate delivery systems. Chitosan was co-administered with polymer microspheres. It was demonstrated that this strategy facilitates markedly enhanced immune responses in both magnitude and duration following intramuscular administration. We conclude that this combination shows potential for single dose administration of vaccines. In another study, we have shown that the addition of chitosan to alum adsorbed TT was able to enhance immune responses. PLA micro/nanospheres were prepared and characterised with discreet particle size ranges. A smaller particle size was shown to facilitate higher scrum IgG responses following nasal administration. A lower antigen loading was additionally identified as being preferential for the induction of immune responses in combination with the smaller particle size. This may be due to the fact that the number of particles will be increased when antigen loading is low, which may in turn facilitate a more widespread uptake of particles. PLA lamellar particles were prepared and characterised. Adsorbed TT was evaluated for the potential to engender immune responses in vivo. These formulations were shown to generate effective immune responses following intramuscular administration. Positively charged polyethylcyanoacrylate and PLA nanoparticies were designed and characterised and their potential as delivery vehicles for DNA vaccines was investigated. Successful preparation of particles with narrow size distribution and positive surface charge (imparted by the inclusion of chitosan) was achieved. In the evaluation of antibody responses to DNA encoded antigen in the presence of alum administered intranasally, discrimination between the groups was only seen following intramuscular boosting with the corresponding protein. Our study showed that DNA vaccines in the presence of either alum or Quil-A may advantageously influence priming of the immune system by a mucosal route. The potential for the combination of adjuvants, Quil-A and chitosan, to enhance antibody responses to plasmid encoded antigen co-administered with the corresponding protein antigen was shown and this is worthy of further investigation. The findings here have identified novel adjuvants and approaches to vaccine delivery. In particular, chitosan or vitamin E TPGS are shown here to have considerable promise as non-toxic, safe mucosal adjuvants. In addition, biodegradable mucoadhesive delivery systems, surface modified with chitosan in a single step process, may have application for other uses such as drug and gene delivery.

Web strength of rolled steel joists

This work is concerned with the behaviour of thin webbed rolled steel joists or universal beams when they are subjected to concentrated loads applied to the flanges. The prime concern is the effect of high direct stresses causing web failure in a small region of the beam. The review shows that although many tests have been carried out on rolled steel beams and built up girders, no series of tests has restricted the number of variables involved to enable firm conclusions to be drawn. The results of 100 tests on several different rolled steel universal beam sections having various types of loading conditions are presented. The majority of the beams are tested by loading with two opposite loads, thus eliminating the effects of bending and shear, except for a small number of beams which are tested simply supported on varying spans. The test results are first compared with the present design standard (BS 449) and it is shown that the British Standard is very conservative for most of the loading conditions included in the tests but is unsafe for others. Three possible failure modes are then considered, overall elastic buckling of the web, flexural yielding of the web due to large out of plane deflexions and local crushing of the material at the junction of the web and the root fillets. Each mode is considered theoretically and developed to establish the main variables, thus enabling a comparison to be made with the test results. It is shown that all three failure modes have a particular relevance for individual loading conditions, but that determining the failure load given the beam size and the loading conditions is very difficult in certain instances. Finally it is shown that there are some empirical relationships between the failure loads and the type of loading for various beam serial sizes.

Examination variability in short-wavelength automated perimetry

The study evaluated sources of within- and between-subject variability in standard white-on-white (W-W) perimetry and short-wavelength automated perimetry (SWAP). The Influence of staircase strategy on the fatigue effect in W-W perimetry was investigated for a 4 dB single step, single reversal strategy; a variable step size, single reversal dynamic strategy; and the standard 4-2 dB double reversal strategy. The fatigue effect increased as the duration of the examination Increased and was greatest in the second eye for all strategies. The fatigue effect was lowest for the 4dB strategy, which exhibited the shortest examination time and was greatest for the 4-2 dB strategy, which exhibited the longest examination time. Staircase efficiency was lowest for the 4 dB strategy and highest for the dynamic strategy which thus offers a reduced examination time and low inter-subject variability. The normal between-subject variability of SWAP was determined for the standard 4-2 dB double reversal strategy and the 3 dB single reversal FASTPAC strategy and compared to that of W-W perimetry, The decrease in sensitivity with Increase in age was greatest for SWAP. The between-subject variability of SWAP was greater than W-W perimetry. Correction for the Influence of ocular media absorption reduced the between-subject variability of SWAP, The FASTPAC strategy yielded the lowest between-subject variability In SWAP, but the greatest between-subject variability In WoW perimetry. The greater between-subject variability of SWAP has profound Implications for the delineation of visual field abnormality, The fatigue effect for the Full Threshold strategy in SWAP was evaluated with conventional opaque, and translucent occlusion of the fellow eye. SWAP exhibited a greater fatigue effect than W-W perimetry. Translucent occlusion reduced the between-subject variability of W-W perimetry but Increased the between-subject variability of SWAP. The elevation of sensitivity was greater with translucent occlusion which has implications for the statistical analysis of W-W perimetry and SWAP. The influence of age-related cataract extraction and IOL implantation upon the visual field derived by WoW perimetry and SWAP was determined. Cataract yielded a general reduction In sensitivity which was preferentially greater in SWAP, even after the correction of SWAP for the attenuation of the stimulus by the ocular media. There was no correlation between either backward or forward light scatter and the magnitude of the attenuation of W-W or SWAP sensitivity. The post-operative mean deviation in SWAP was positive and has ramifications for the statistical Interpretation of SWAP. Short-wavelength-sensitive pathway isolation was assessed as a function of stimulus eccentricity using the two-colour Increment threshold method. At least 15 dB of SWS pathway Isolation was achieved for 440 nm, 450 nm and 460 nm stimuli at a background luminance of 100 cdm-2, There was a slight decrease In SWS pathway Isolation for all stimulus wavelengths with increasing eccentricity which was not of clinical significance. Adopting a 450 nm stimulus may reduce between-subject variability In SWAP due to a reduction In ocular media absorption and macular pigment absorption.

The dynamics and control of a remotely piloted plan-symmetric helicopter

Prior to the development of a production standard control system for ML Aviation's plan-symmetric remotely piloted helicopter system, SPRITE, optimum solutions to technical requirements had yet to be found for some aspects of the work. This thesis describes an industrial project where solutions to real problems have been provided within strict timescale constraints. Use has been made of published material wherever appropriate, new solutions have been contributed where none existed previously. A lack of clearly defined user requirements from potential Remotely Piloted Air Vehicle (RPAV) system users is identified, A simulation package is defined to enable the RPAV designer to progress with air vehicle and control system design, development and evaluation studies and to assist the user to investigate his applications. The theoretical basis of this simulation package is developed including Co-axial Contra-rotating Twin Rotor (CCTR), six degrees of freedom motion, fuselage aerodynamics and sensor and control system models. A compatible system of equations is derived for modelling a miniature plan-symmetric helicopter. Rigorous searches revealed a lack of CCTR models, based on closed form expressions to obviate integration along the rotor blade, for stabilisation and navigation studies through simulation. An economic CCTR simulation model is developed and validated by comparison with published work and practical tests. Confusion in published work between attitude and Euler angles is clarified. The implementation of package is discussed. dynamic adjustment of assessment. the theory into a high integrity software Use is made of a novel technique basing the integration time step size on error Simulation output for control system stability verification, cross coupling of motion between control channels and air vehicle response to demands and horizontal wind gusts studies are presented. Contra-Rotating Twin Rotor Flight Control System Remotely Piloted Plan-Symmetric Helicopter Simulation Six Degrees of Freedom Motion ( i i)

What determines the size frequency distribution of β-amyloid (Aβ) deposits in Alzheimer's disease patients?

The factors determining the size of individual β-amyloid (A,8) deposits and their size frequency distribution in tissue from Alzheimer's disease (AD) patients have not been established. In 23/25 cortical tissues from 10 AD patients, the frequency of Aβ deposits declined exponentially with increasing size. In a random sample of 400 Aβ deposits, 88% were closely associated with one or more neuronal cell bodies. The frequency distribution of (Aβ) deposits which were associated with 0,1,2,...,n neuronal cell bodies deviated significantly from a Poisson distribution, suggesting a degree of clustering of the neuronal cell bodies. In addition, the frequency of Aβ deposits declined exponentially as the number of associated neuronal cell bodies increased. Aβ deposit area was positively correlated with the frequency of associated neuronal cell bodies, the degree of correlation being greater for pyramidal cells than smaller neurons. These data suggested: (1) the number of closely adjacent neuronal cell bodies which simultaneously secrete Aβ was an important factor determining the size of an Aβ deposit and (2) the exponential decline in larger Aβ deposits reflects the low probability that larger numbers of adjacent neurons will secrete Aβ simultaneously to form a deposit. © 1995.