Bayesian invariant measurements of generalisation

The problem of evaluating different learning rules and other statistical estimators is analysed. A new general theory of statistical inference is developed by combining Bayesian decision theory with information geometry. It is coherent and invariant. For each sample a unique ideal estimate exists and is given by an average over the posterior. An optimal estimate within a model is given by a projection of the ideal estimate. The ideal estimate is a sufficient statistic of the posterior, so practical learning rules are functions of the ideal estimator. If the sole purpose of learning is to extract information from the data, the learning rule must also approximate the ideal estimator. This framework is applicable to both Bayesian and non-Bayesian methods, with arbitrary statistical models, and to supervised, unsupervised and reinforcement learning schemes.

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.

Profitability of a sample of Portuguese bank branches and its decomposition into technical and allocative components

The efficiency literature, both using parametric and non-parametric methods, has been focusing mainly on cost efficiency analysis rather than on profit efficiency. In for-profit organisations, however, the measurement of profit efficiency and its decomposition into technical and allocative efficiency is particularly relevant. In this paper a newly developed method is used to measure profit efficiency and to identify the sources of any shortfall in profitability (technical and/or allocative inefficiency). The method is applied to a set of Portuguese bank branches first assuming long run and then a short run profit maximisation objective. In the long run most of the scope for profit improvement of bank branches is by becoming more allocatively efficient. In the short run most of profit gain can be realised through higher technical efficiency. © 2003 Elsevier B.V. All rights reserved.

The missing link:Analogous human and primate cortical gamma oscillations

Recent animal studies highlighting the relationship between functional imaging signals and the underlying neuronal activity have revealed the potential capabilities of non-invasive methods. However, the valuable exchange of information between animal and human studies remains restricted by the limited evidence of direct physiological links between species. In this study we used magnetoencephalography (MEG) to investigate the occurrence of 30-70 Hz (gamma) oscillations in human visual cortex, induced by the presentation of visual stimuli of varying contrast. These oscillations, well described in the animal literature, were observed in retinotopically concordant locations of visual cortex and show striking similarity to those found in primate visual cortex using surgically implanted electrodes. The amplitude of the gamma oscillations increases linearly with stimulus contrast in strong correlation with the gamma oscillations found in the local field potential (LFP) of the macaque. We demonstrate that non-invasive magnetic field measurements of gamma oscillations in human visual cortex concur with invasive measures of activation in primate visual cortex, suggesting both a direct representation of underlying neuronal activity and a concurrence between human and primate cortical activity. © 2005 Elsevier Inc. All rights reserved.

Lipoidal species in ocular spoilation processes

Tear component deposition onto contact lenses is termed `spoilation' and occurs due to the interaction of synthetic polymers with their biological fluid environment. Spoilation phenomena alter the physico-chemical properties of hydrophilic contact lenses, diminishing the optical properties of the lens; causing discomfort and complications for the wearer. Eventually these alterations render the lens unwearable. The primary aim of this interdisciplinary study was to develop analytical techniques capable of analysing the minute quantities of biological deposition involved, in particular the lipid fraction. Prior to this work such techniques were unavailable for single contact lenses. It is envisaged that these investigations will further the understanding of this biological interfacial conversion. Two main analytical techniques were developed: a high performance liquid chromatography (HPLC) technique and fluorescence spectrofluorimetry. The HPLC method allows analysis of a single contact lens and provided previously unavailable valuable information about variations in the lipid profiles of deposited contact lenses and patient tear films. Fluorescence spectrophotofluorimetry is a sensitive non-destructive technique for observing changes in the fluorescence intensity of biological components on contact lenses. The progression and deposition of tear materials can be monitored and assessed for both in vivo and in vitro spoiled lenses using this technique. An improved in vitro model which is comparable to tears and chemically mimics ocular spoilation was also developed. This model allows the controlled study of extrinsic factors and hydrogel compositions. These studies show that unsaturated tear lipids, probably unsaturated fatty acids, are involved in the interfacial conversion of hydrogel lenses, rendering them incompatible with the ocular microenvironment. Lipid interaction with the lens surface then facilitates secondary deposition of other tear components. Interaction, exchange and immobilisation (by polymerisation) of the lipid layer appears to occur before the final and rapid growth of more complex, insoluble discrete deposits, sometimes called `white spots'.

Composite materials impact damage detection using neural networks

This thesis considers two basic aspects of impact damage in composite materials, namely damage severity discrimination and impact damage location by using Acoustic Emissions (AE) and Artificial Neural Networks (ANNs). The experimental work embodies a study of such factors as the application of AE as Non-destructive Damage Testing (NDT), and the evaluation of ANNs modelling. ANNs, however, played an important role in modelling implementation. In the first aspect of the study, different impact energies were used to produce different level of damage in two composite materials (T300/914 and T800/5245). The impacts were detected by their acoustic emissions (AE). The AE waveform signals were analysed and modelled using a Back Propagation (BP) neural network model. The Mean Square Error (MSE) from the output was then used as a damage indicator in the damage severity discrimination study. To evaluate the ANN model, a comparison was made of the correlation coefficients of different parameters, such as MSE, AE energy, AE counts, etc. MSE produced an outstanding result based on the best performance of correlation. In the second aspect, a new artificial neural network model was developed to provide impact damage location on a quasi-isotropic composite panel. It was successfully trained to locate impact sites by correlating the relationship between arriving time differences of AE signals at transducers located on the panel and the impact site coordinates. The performance of the ANN model, which was evaluated by calculating the distance deviation between model output and real location coordinates, supports the application of ANN as an impact damage location identifier. In the study, the accuracy of location prediction decreased when approaching the central area of the panel. Further investigation indicated that this is due to the small arrival time differences, which defect the performance of ANN prediction. This research suggested increasing the number of processing neurons in the ANNs as a practical solution.

Sensing and control of adipocyte function

Obesity has become a global epidemic. Approximately 15% of the world population is either overweight or obese. This figure rises to 75% in many westernised countries including the United Kingdom. Health costs in the UK to treat obesity and associated disease are conservatively estimated at 6% of the National Health Service (NHS) budget equating to 3.33 billion Euros. Excess adiposity, especially in visceral depots, increases the risk of type 2 diabetes, cardiovascular disease, gall stones, hypertension and cancer. Type 2 diabetes mellitus accounts for >90% of all cases of diabetes of which the majority can be attributed to increased adiposity, and approximately 70% of cardiovascular disease has been attributed to obesity in the US. Weight loss reduces risk of these complications and in some cases can eliminate the condition. However, weight loss by conventional non-medicated methods is often unsuccessful or promptly followed by weight regain. This thesis has investigated adipocytes development and adipokine signalling with a view to enhance the understanding of tissue functionality and to identify possible targets or pathways for therapeutic intervention. Adipocyte isolation from human tissue samples was undertaken for these investigative studies, and the methodology was optimised. The resulting isolates of pre-adipocytes and mature adipocytes were characterised and evaluated. Major findings from these studies indicate that mature adipocytes undergo cell division post terminal differentiation. Gene studies indicated that subcutaneous adipose tissue exuded greater concentrations and fluctuations of adipokine levels than visceral adipose tissue, indicating an important adiposensing role of subcutaneous adipose tissue. It was subsequently postulated that the subcutaneous depot may provide the major focus for control of overall energy balance and by extension weight control. One potential therapeutic target, 11ß-hydrosteroid dehydrogenase (11ß-HSD1) was investigated, and prospective inhibitors of its action were considered (BVT1, BVT2 and AZ121). Selective reduction of adiposity of the visceral depot was desired due to its correlation with the detrimental effects of obesity. However, studies indicated that although the visceral depot tissue was not unaffected, the subcutaneous depot was more susceptible to therapeutic inhibition by these compounds. This was determined to be a potentially valuable therapeutic intervention in light of previous postulations regarding long-term energy control via the subcutaneous tissue depot.

Suitability of dysphonia measurements for telemonitoring of Parkinson's disease

We present an assessment of the practical value of existing traditional and non-standard measures for discriminating healthy people from people with Parkinson's disease (PD) by detecting dysphonia. We introduce a new measure of dysphonia, Pitch Period Entropy (PPE), which is robust to many uncontrollable confounding effects including noisy acoustic environments and normal, healthy variations in voice frequency. We collected sustained phonations from 31 people, 23 with PD. We then selected 10 highly uncorrelated measures, and an exhaustive search of all possible combinations of these measures finds four that in combination lead to overall correct classification performance of 91.4%, using a kernel support vector machine. In conclusion, we find that non-standard methods in combination with traditional harmonics-to-noise ratios are best able to separate healthy from PD subjects. The selected non-standard methods are robust to many uncontrollable variations in acoustic environment and individual subjects, and are thus well-suited to telemonitoring applications.

Development of an enantiomer selective microsampling assay for the quantification of ketorolac suitable for paediatric pharmacokinetic studies

Background: Ketorolac, a potent nonsteroidal anti-inflammatory drug used for pain control in children, exists as a racemate of inactive R (+) and active S (-) enantiomers. Aim: To develop a microsampling assay for the enantioselective analysis of ketorolac in children. Methods: Ketorolac enantiomers were extracted from 50 µl of plasma by liquid–liquid extraction and separated on a ChiralPak AD-RH. Detection was by a TSQ quantum triple quadrupole mass spectrometer with an electrospray ionisation source operating in a positive ion mode. Five children (age 13.8 (1.6) years, weight 52.7 (7.2) kg), were administered intravenous ketorolac 0.5 mg/kg (maximum 10 mg) and blood samples were taken at 0, 0.25, 0.5, 1, 2, 4, 6, 8 and 12 h post administration. CL, VD and t1/2 were calculated based on non-compartmental methods. Results: The standard curves for R (+) and S (-) ketorolac were linear in the range 0–2000 ng/ml. The LLOQs of the method were 0.15 ng on column and 0.31 ng on column for R (+) and S (-) ketorolac, respectively. The median (range) VD and CL of R (+) and S (-) ketorolac were 0.12 l/kg (0.07–0.17), 0.017 l/h/kg (0.12–0.29) and 0.17 (0.09–0.31) l/kg, 0.049 (0.02–0.1) l/h/kg, p = 0.043), respectively. The median (range) elimination half-life (t1/2) of the R (+) and S (-) ketorolac was 5.0 h (2.5–5.8) and 3.1 h (1.8–4.4), p = 0.043), respectively. Conclusion: The development of a simple, rapid and reliable ketorolac assay suitable for paediatric PK studies is reported. Copyright © 2013 John Wiley & Sons, Ltd.

Customer loyalty among daily disposable contact lens wearers

Background: Optometric practices offer contact lenses as cash sale items or as part of monthly payment plans. With the contact lens market becoming increasingly competitive, patients are opting to purchase lenses from supermarkets and Internet suppliers. Monthly payment plans are often implemented to improve loyalty. This study aimed to compare behavioural loyalty between monthly payment plan members and non-members. Methods: BBR Optometry Ltd offers a monthly payment plan (Eyelife™) to their contact lens wearers. A retrospective audit of 38 Eyelife™ members (mean. ±. SD: 42.7. ±. 15.0 years) and 30 non-members (mean. ±. SD: 40.8. ±. 16.7 years) was conducted. Revenue and profits generated, service uptake and product sales between the two groups were compared over a fixed period of 18 months. Results: Eyelife™ members generated significantly higher professional fee revenue ( P<. 0.001), £153.96 compared to £83.50, and profits ( P<. 0.001). Eyelife™ members had a higher uptake of eye examinations ( P<. 0.001). The 2 groups demonstrated no significant difference in spectacle sales by volume ( P= 0.790) or value ( P= 0.369). There were also no significant differences in contact lens revenue ( P= 0.337), although Eyelife™ members did receive a discount. The Eyelife™ group incurred higher contact lens costs ( P= 0.037), due to a greater volume of contact lens purchases, 986 units compared to 582. Conclusions: Monthly payment plans improve loyalty among contact lens wearers, particularly service uptake and volume of lens purchases. Additionally the greater professional fees generated, render monthly payment plans an attractive business model and practice builder.

Why is it possible to detect doped regions of semiconductors in low voltage SEM:a review and update

There is an urgent need for fast, non-destructive and quantitative two-dimensional dopant profiling of modern and future ultra large-scale semiconductor devices. The low voltage scanning electron microscope (LVSEM) has emerged to satisfy this need, in part, whereby it is possible to detect different secondary electron yield values (brightness in the SEM signal) from the p-type to the n-type doped regions as well as different brightness levels from the same dopant type. The mechanism that gives rise to such a secondary electron (SE) contrast effect is not fully understood, however. A review of the different models that have been proposed to explain this SE contrast is given. We report on new experiments that support the proposal that this contrast is due to the establishment of metal-to-semiconductor surface contacts. Further experiments showing the effect of instrument parameters including the electron dose, the scan speeds and the electron beam energy on the SE contrast are also reported. Preliminary results on the dependence of the SE contrast on the existence of a surface structure featuring metal-oxide semiconductor (MOS) are also reported. Copyright © 2005 John Wiley & Sons, Ltd.

Applications of data envelopment analysis in education

Non-parametric methods for efficiency evaluation were designed to analyse industries comprising multi-input multi-output producers and lacking data on market prices. Education is a typical example. In this chapter, we review applications of DEA in secondary and tertiary education, focusing on the opportunities that this offers for benchmarking at institutional level. At secondary level, we investigate also the disaggregation of efficiency measures into pupil-level and school-level effects. For higher education, while many analyses concern overall institutional efficiency, we examine also studies that take a more disaggregated approach, centred either around the performance of specific functional areas or that of individual employees.

Analysis of Hue Spectra During Storage of Pepper (Capsicum annuum)

Color information is widely used in non-destructive quality assessment of perishable horticultural produces. The presented work investigated color changes of pepper (Capsicum annuum L.) samples received from retail system. The effect of storage temperature (10±2°C and 24±4°C) on surface color and firmness was analyzed. Hue spectra was calculated using sum of saturations. A ColorLite sph850 (400-700nm) spectrophotometer was used as reference instrument. Dynamic firmness was measured on three locations of the surface: tip cap, middle and shoulder. Significant effects of storage conditions and surface location on both color and firmness were observed. Hue spectra responded sensitively to color development of pepper. Prediction model (PLS) was used to estimate dynamic firmess based on hue spectra. Accuracy was very different depending on the location. Firmness of the tip cap was predicted with the highest accuracy (RMSEP=0.0335). On the other hand, middle region cannot be used for such purpose. Due to the simplicity and rapid processing, analysis of hue spectra is a promising tool for evaluation of color in postharvest and food industry.

Segmental body composition determination through a bioelectrical very large scale integration (VLSI) system

The purpose of this investigation was to develop new techniques to generate segmental assessments of body composition based on Segmental Bioelectrical Impedance Analysis (SBIA). An equally important consideration was the design, simulation, development, and the software and hardware integration of the SBIA system. This integration was carried out with a Very Large Scale Integration (VLSI) Field Programmable Gate Array (FPGA) microcontroller that analyzed the measurements obtained from segments of the body, and provided full body and segmental Fat Free Mass (FFM) and Fat Mass (FM) percentages. Also, the issues related to the estimate of the body's composition in persons with spinal cord injury (SCI) were addressed and investigated. This investigation demonstrated that the SBIA methodology provided accurate segmental body composition measurements. Disabled individuals are expected to benefit from these SBIA evaluations, as they are non-invasive methods, suitable for paralyzed individuals. The SBIA VLSI system may replace bulky, non flexible electronic modules attached to human bodies. ^

Responses of sawgrass and spikerush to variation in hydrologic drivers and salinity in Southern Everglades marshes

Aboveground net primary production (ANPP) by the dominant macrophyte and plant community composition are related to the changing hydrologic environment and to salinity in the southern Everglades, FL, USA. We present a new non-destructive ANPP technique that is applicable to any continuously growing herbaceous system. Data from 16 sites, collected from 1998 to 2004, were used to investigate how hydrology and salinity controlled sawgrass (Cladium jamaicense Crantz.) ANPP. Sawgrass live biomass showed little seasonal variation and annual means ranged from 89 to 639 gdw m)2. Mortality rates were 20–35% of live biomass per 2 month sampling interval, for biomass turnover rates of 1.3–2.5 per year. Production by C. jamaicense was manifest primarily as biomass turnover, not as biomass accumulation. Rates typically ranged from 300 to 750 gdw m)2 year)1, but exceeded 1000 gdw m)2 year)1 at one site and were as high as 750 gdw m)2 year)1 at estuarine ecotone sites. Production was negatively related to mean annual water depth, hydroperiod, and to a variable combining the two (depth-days). As water depths and hydroperiods increased in our southern Everglades study area, sawgrass ANPP declined. Because a primary restoration goal is to increase water depths and hydroperiods for some regions of the Everglades, we investigated how the plant community responded to this decline in sawgrass ANPP. Spikerush (Eleocharis sp.) was the next most prominent component of this community at our sites, and 39% of the variability in sawgrass ANPP was explained by a negative relationship with mean annual water depth, hydroperiod, and Eleocharis sp. density the following year. Sawgrass ANPP at estuarine ecotone sites responded negatively to salinity, and rates of production were slow to recover after high salinity years. Our results suggest that ecologists, managers, and the public should not necessarily interpret a decline in sawgrass that may result from hydrologic restoration as a negative phenomenon.