Training with noise is equivalent to Tikhonov regularization

It is well known that the addition of noise to the input data of a neural network during training can, in some circumstances, lead to significant improvements in generalization performance. Previous work has shown that such training with noise is equivalent to a form of regularization in which an extra term is added to the error function. However, the regularization term, which involves second derivatives of the error function, is not bounded below, and so can lead to difficulties if used directly in a learning algorithm based on error minimization. In this paper we show that, for the purposes of network training, the regularization term can be reduced to a positive definite form which involves only first derivatives of the network mapping. For a sum-of-squares error function, the regularization term belongs to the class of generalized Tikhonov regularizers. Direct minimization of the regularized error function provides a practical alternative to training with noise.

Mathematical modelling and optimal multivariable control of chemical processes

This work reports the developnent of a mathenatical model and distributed, multi variable computer-control for a pilot plant double-effect climbing-film evaporator. A distributed-parameter model of the plant has been developed and the time-domain model transformed into the Laplace domain. The model has been further transformed into an integral domain conforming to an algebraic ring of polynomials, to eliminate the transcendental terms which arise in the Laplace domain due to the distributed nature of the plant model. This has made possible the application of linear control theories to a set of linear-partial differential equations. The models obtained have well tracked the experimental results of the plant. A distributed-computer network has been interfaced with the plant to implement digital controllers in a hierarchical structure. A modern rnultivariable Wiener-Hopf controller has been applled to the plant model. The application has revealed a limitation condition that the plant matrix should be positive-definite along the infinite frequency axis. A new multi variable control theory has emerged fram this study, which avoids the above limitation. The controller has the structure of the modern Wiener-Hopf controller, but with a unique feature enabling a designer to specify the closed-loop poles in advance and to shape the sensitivity matrix as required. In this way, the method treats directly the interaction problems found in the chemical processes with good tracking and regulation performances. Though the ability of the analytical design methods to determine once and for all whether a given set of specifications can be met is one of its chief advantages over the conventional trial-and-error design procedures. However, one disadvantage that offsets to some degree the enormous advantages is the relatively complicated algebra that must be employed in working out all but the simplest problem. Mathematical algorithms and computer software have been developed to treat some of the mathematical operations defined over the integral domain, such as matrix fraction description, spectral factorization, the Bezout identity, and the general manipulation of polynomial matrices. Hence, the design problems of Wiener-Hopf type of controllers and other similar algebraic design methods can be easily solved.

Studies on the generalisation of Gaussian processes and Bayesian neural networks

The assessment of the reliability of systems which learn from data is a key issue to investigate thoroughly before the actual application of information processing techniques to real-world problems. Over the recent years Gaussian processes and Bayesian neural networks have come to the fore and in this thesis their generalisation capabilities are analysed from theoretical and empirical perspectives. Upper and lower bounds on the learning curve of Gaussian processes are investigated in order to estimate the amount of data required to guarantee a certain level of generalisation performance. In this thesis we analyse the effects on the bounds and the learning curve induced by the smoothness of stochastic processes described by four different covariance functions. We also explain the early, linearly-decreasing behaviour of the curves and we investigate the asymptotic behaviour of the upper bounds. The effect of the noise and the characteristic lengthscale of the stochastic process on the tightness of the bounds are also discussed. The analysis is supported by several numerical simulations. The generalisation error of a Gaussian process is affected by the dimension of the input vector and may be decreased by input-variable reduction techniques. In conventional approaches to Gaussian process regression, the positive definite matrix estimating the distance between input points is often taken diagonal. In this thesis we show that a general distance matrix is able to estimate the effective dimensionality of the regression problem as well as to discover the linear transformation from the manifest variables to the hidden-feature space, with a significant reduction of the input dimension. Numerical simulations confirm the significant superiority of the general distance matrix with respect to the diagonal one.In the thesis we also present an empirical investigation of the generalisation errors of neural networks trained by two Bayesian algorithms, the Markov Chain Monte Carlo method and the evidence framework; the neural networks have been trained on the task of labelling segmented outdoor images.

Ocular and systemic vascular function in human ageing

The importance of vascular risk factors in various age-related pathologies has been extensively researched. Nevertheless, the haemodynamic disturbance occurring in various ocular and systemic vascular beds to impact upon ocular function remained largely unknown. The purpose of the following studies was to explore the presence and impact of both ocular and systemic vascular dysregulation as well as biochemical vascular risk factors in healthy elderly individuals and patients with age-related macular degeneration. Furthermore, the possible role was played by circulatory oxidative stress and its relationship with endothelial dysfunction at both ocular and systemic levels has also been investigated. There were four principal sections to the present work: 1. To assess the relationship between ocular and systemic anti-oxidative defence in healthy individuals The principal findings of this work were: -It has been shown that MPOD significantly and positively related with circulatory GSH levels. 2. To investigate macro- and microcirculation and oxidative stress in early AMD patients without overt systemic disease The principal findings of this work were: -AMD patients exhibit abnormal macrocirculation compared to the controls. -Blood GSSG level was significantly higher in early AMD patients than controls. -AMD patients showed abnormal microcirculation at retinal level compared. -In early AMD patients, retinal venous RT positively correlated with blood GSSG levels. 3. To assess the relationship between ocular vascular function and circulatory markers of endothelial dysfunction and CVD risk The principal findings of this work were: -Age had a positive effect on ET-1, vWF and slope of retinal arterial constriction in otherwise healthy individuals. -Even in otherwise healthy individuals, retinal arterial vascular function showed a significant correlation with circulatory markers of endothelial dysfunction and CVD risk. 4. To assess age-related changes in ANS and vascular function, and their relationship to retinal vascular function parameters The principal findings of this work were: -Elderly individuals demonstrated abnormal circadian changes of PSNS activity compared to middle-aged group. -Elderly groups showed higher ET-1 and vWF level as well as C-IMT and AIx, and also impaired retinal vascular function compared to the middle-aged group. -In the elderly group, impaired retinal vascular function significantly correlated with the dysregulation of PSNS activity.

Anticholinergic medication use and cognitive impairment in the older population:the Medical Research Council cognitive function and ageing study

OBJECTIVES: To determine whether the use of medications with possible and definite anticholinergic activity increases the risk of cognitive impairment and mortality in older people and whether risk is cumulative. DESIGN: A 2-year longitudinal study of participants enrolled in the Medical Research Council Cognitive Function and Ageing Study between 1991 and 1993. SETTING: Community-dwelling and institutionalized participants. PARTICIPANTS: Thirteen thousand four participants aged 65 and older. MEASUREMENTS: Baseline use of possible or definite anticholinergics determined according to the Anticholinergic Cognitive Burden Scale and cognition determined using the Mini-Mental State Examination (MMSE). The main outcome measure was decline in the MMSE score at 2 years. RESULTS: At baseline, 47% of the population used a medication with possible anticholinergic properties, and 4% used a drug with definite anticholinergic properties. After adjusting for age, sex, educational level, social class, number of nonanticholinergic medications, number of comorbid health conditions, and cognitive performance at baseline, use of medication with definite anticholinergic effects was associated with a 0.33-point greater decline in MMSE score (95% confidence interval (CI)=0.03–0.64, P=.03) than not taking anticholinergics, whereas the use of possible anticholinergics at baseline was not associated with further decline (0.02, 95% CI=-0.14–0.11, P=.79). Two-year mortality was greater for those taking definite (OR=1.68; 95% CI=1.30–2.16; P<.001) and possible (OR=1.56; 95% CI=1.36–1.79; P<.001) anticholinergics. CONCLUSION: The use of medications with anticholinergic activity increases the cumulative risk of cognitive impairment and mortality.

Executive function and emotional focus in autobiographical memory specificity in older adults

The current study examined the role of executive function in retrieval of specific autobiographical memories in older adults with regard to control of emotion during retrieval. Older and younger adults retrieved memories of specific events in response to emotionally positive, negative and neutral word cues. Contributions of inhibitory and updating elements of executive function to variance in autobiographical specificity were assessed to determine processes involved in the commonly found age-related reduction in specificity. A negative relationship between age and specificity was only found in retrieval to neutral cues. Alternative explanations of this age preservation of specificity of emotional recall are explored, within the context of control of emotion in the self-memory system and preserved emotional processing and positivity effect in older adults. The pattern of relationships suggests updating, rather than inhibition as the source of age-related reduction in specificity, but that emotional processing (particularly of positively valenced memories) is not influenced by age-related variance in executive control. The tendency of older adults to focus on positive material may thus act as a buffer against detrimental effects of reduced executive function capacity on autobiographical retrieval, representing a possible target for interventions to improve specificity of autobiographical memory retrieval in older adults.

A randomised controlled trial investigating the effect of nutritional supplementation on visual function in normal, and age-related macular disease affected eyes:design and methodology [ISRCTN78467674]

Background: Age-related macular disease is the leading cause of blind registration in the developed world. One aetiological hypothesis involves oxidation, and the intrinsic vulnerability of the retina to damage via this process. This has prompted interest in the role of antioxidants, particularly the carotenoids lutein and zeaxanthin, in the prevention and treatment of this eye disease. Methods: The aim of this randomised controlled trial is to determine the effect of a nutritional supplement containing lutein, vitamins A, C and E, zinc, and copper on measures of visual function in people with and without age-related macular disease. Outcome measures are distance and near visual acuity, contrast sensitivity, colour vision, macular visual field, glare recovery, and fundus photography. Randomisation is achieved via a random number generator, and masking achieved by third party coding of the active and placebo containers. Data collection will take place at nine and 18 months, and statistical analysis will employ Student's t test. Discussion: A paucity of treatment modalities for age-related macular disease has prompted research into the development of prevention strategies. A positive effect on normals may be indicative of a role of nutritional supplementation in preventing or delaying onset of the condition. An observed benefit in the age-related macular disease group may indicate a potential role of supplementation in prevention of progression, or even a degree reversal of the visual effects caused by this condition.

Nutritional influences on cognitive function:mechanisms of susceptibility

The impact of nutritional variation, within populations not overtly malnourished, on cognitive function and arousal is considered. The emphasis is on susceptibility to acute effects of meals and glucose loads, and chronic effects of dieting, on mental performance, and effects of cholesterol and vitamin levels on cognitive impairment. New developments in understanding dietary influences on neurohormonal systems, and their implications for cognition and affect, allow reinterpretation of both earlier and recent findings. Evidence for a detrimental effect of omitting a meal on cognitive performance remains equivocal: from the outset, idiosyncrasy has prevailed. Yet, for young and nutritionally vulnerable children, breakfast is more likely to benefit than hinder performance. For nutrient composition, despite inconsistencies, some cautious predictions can be made. Acutely, carbohydrate-rich–protein-poor meals can be sedating and anxiolytic; by comparison, protein-rich meals may be arousing, improving reaction time but also increasing unfocused vigilance. Fat-rich meals can lead to a decline in alertness, especially where they differ from habitual fat intake. These acute effects may vary with time of day and nutritional status. Chronically, protein-rich diets have been associated with decreased positive and increased negative affect relative to carbohydrate-rich diets. Probable mechanisms include diet-induced changes in monoamine, especially serotoninergic neurotransmitter activity, and functioning of the hypothalamic pituitary adrenal axis. Effects are interpreted in the context of individual traits and susceptibility to challenging, even stressful, tests of performance. Preoccupation with dieting may impair cognition by interfering with working memory capacity, independently of nutritional status. The change in cognitive performance after administration of glucose, and other foods, may depend on the level of sympathetic activation, glucocorticoid secretion, and pancreatic β-cell function, rather than simple fuelling of neural activity. Thus, outcomes can be predicted by vulnerability in coping with stressful challenges, interacting with nutritional history and neuroendocrine status. Functioning of such systems may be susceptible to dietary influences on neural membrane fluidity, and vitamin-dependent cerebrovascular health, with cognitive vulnerability increasing with age.