Time delay estimation with hidden Markov models

Most traditional methods for extracting the relationships between two time series are based on cross-correlation. In a non-linear non-stationary environment, these techniques are not sufficient. We show in this paper how to use hidden Markov models to identify the lag (or delay) between different variables for such data. Adopting an information-theoretic approach, we develop a procedure for training HMMs to maximise the mutual information (MMI) between delayed time series. The method is used to model the oil drilling process. We show that cross-correlation gives no information and that the MMI approach outperforms maximum likelihood.

Hidden state models for time series

Amongst all the objectives in the study of time series, uncovering the dynamic law of its generation is probably the most important. When the underlying dynamics are not available, time series modelling consists of developing a model which best explains a sequence of observations. In this thesis, we consider hidden space models for analysing and describing time series. We first provide an introduction to the principal concepts of hidden state models and draw an analogy between hidden Markov models and state space models. Central ideas such as hidden state inference or parameter estimation are reviewed in detail. A key part of multivariate time series analysis is identifying the delay between different variables. We present a novel approach for time delay estimating in a non-stationary environment. The technique makes use of hidden Markov models and we demonstrate its application for estimating a crucial parameter in the oil industry. We then focus on hybrid models that we call dynamical local models. These models combine and generalise hidden Markov models and state space models. Probabilistic inference is unfortunately computationally intractable and we show how to make use of variational techniques for approximating the posterior distribution over the hidden state variables. Experimental simulations on synthetic and real-world data demonstrate the application of dynamical local models for segmenting a time series into regimes and providing predictive distributions.

Development of an electric heat pump for domestic use

This thesis records the design and development of an electrically driven, air to water, vapour compression heat pump of nominally 6kW heat output, for residential space heating. The study was carried out on behalf of GEC Research Ltd through the Interdisciplinary Higher Degrees Scheme at Aston University. A computer based mathematical model of the vapour compression cycle was produced as a design aid, to enable the effects of component design changes or variations in operating conditions to be predicted. This model is supported by performance testing of the major components, which revealed that improvements in the compressor isentropic efficiency offer the greatest potential for further increases in cycle COPh. The evaporator was designed from first principles, and is based on wire-wound heat transfer tubing. Two evaporators, of air side area 10.27 and 16.24m2, were tested in a temperature and humidity controlled environment, demonstrating that the benefits of the large coil are greater heat pump heat output and lower noise levels. A systematic study of frost growth rates suggested that this problem is most severe at the conditions of saturated air at 0oC combined with low condenser water temperature. A dynamic simulation model was developed to predict the in-service performance of the heat pump. This study confirmed the importance of an adequate radiator area for heat pump installations. A prototype heat pump was designed and manufactured, consisting of a hermetic reciprocating compressor, a coaxial tube condenser and a helically coiled evaporator, using Refrigerant 22. The prototype was field tested in a domestic environment for one and a half years. The installation included a comprehensive monitoring system. Initial problems were encountered with defrosting and compressor noise, both of which were solved. The unit then operated throughout the 1985/86 heating season without further attention, producing a COPh of 2.34.

Multispectral retinal image analysis:a novel non-invasive tool for retinal imaging

Purpose-To develop a non-invasive method for quantification of blood and pigment distributions across the posterior pole of the fundus from multispectral images using a computer-generated reflectance model of the fundus. Methods - A computer model was developed to simulate light interaction with the fundus at different wavelengths. The distribution of macular pigment (MP) and retinal haemoglobins in the fundus was obtained by comparing the model predictions with multispectral image data at each pixel. Fundus images were acquired from 16 healthy subjects from various ethnic backgrounds and parametric maps showing the distribution of MP and of retinal haemoglobins throughout the posterior pole were computed. Results - The relative distributions of MP and retinal haemoglobins in the subjects were successfully derived from multispectral images acquired at wavelengths 507, 525, 552, 585, 596, and 611?nm, providing certain conditions were met and eye movement between exposures was minimal. Recovery of other fundus pigments was not feasible and further development of the imaging technique and refinement of the software are necessary to understand the full potential of multispectral retinal image analysis. Conclusion - The distributions of MP and retinal haemoglobins obtained in this preliminary investigation are in good agreement with published data on normal subjects. The ongoing development of the imaging system should allow for absolute parameter values to be computed. A further study will investigate subjects with known pathologies to determine the effectiveness of the method as a screening and diagnostic tool.

Combining temporal and spectral information with spatial mapping to identify differences between phonological and semantic networks:a magnetoencephalographic approach

Early, lesion-based models of language processing suggested that semantic and phonological processes are associated with distinct temporal and parietal regions respectively, with frontal areas more indirectly involved. Contemporary spatial brain mapping techniques have not supported such clear-cut segregation, with strong evidence of activation in left temporal areas by both processes and disputed evidence of involvement of frontal areas in both processes. We suggest that combining spatial information with temporal and spectral data may allow a closer scrutiny of the differential involvement of closely overlapping cortical areas in language processing. Using beamforming techniques to analyze magnetoencephalography data, we localized the neuronal substrates underlying primed responses to nouns requiring either phonological or semantic processing, and examined the associated measures of time and frequency in those areas where activation was common to both tasks. Power changes in the beta (14-30 Hz) and gamma (30-50 Hz) frequency bandswere analyzed in pre-selected time windows of 350-550 and 500-700ms In left temporal regions, both tasks elicited power changes in the same time window (350-550 ms), but with different spectral characteristics, low beta (14-20 Hz) for the phonological task and high beta (20-30 Hz) for the semantic task. In frontal areas (BA10), both tasks elicited power changes in the gamma band (30-50 Hz), but in different time windows, 500-700ms for the phonological task and 350-550ms for the semantic task. In the left inferior parietal area (BA40), both tasks elicited changes in the 20-30 Hz beta frequency band but in different time windows, 350-550ms for the phonological task and 500-700ms for the semantic task. Our findings suggest that, where spatial measures may indicate overlapping areas of involvement, additional beamforming techniques can demonstrate differential activation in time and frequency domains. © 2012 McNab, Hillebrand, Swithenby and Rippon.

A common rule for integration and suppression of luminance contrast across eyes, space, time, and pattern

Visual perception begins by dissecting the retinal image into millions of small patches for local analyses by local receptive fields. However, image structures extend well beyond these receptive fields and so further processes must be involved in sewing the image fragments back together to derive representations of higher order (more global) structures. To investigate the integration process, we also need to understand the opposite process of suppression. To investigate both processes together, we measured triplets of dipper functions for targets and pedestals involving interdigitated stimulus pairs (A, B). Previous work has shown that summation and suppression operate over the full contrast range for the domains of ocularity and space. Here, we extend that work to include orientation and time domains. Temporal stimuli were 15-Hz counter-phase sine-wave gratings, where A and B were the positive and negative phases of the oscillation, respectively. For orientation, we used orthogonally oriented contrast patches (A, B) whose sum was an isotropic difference of Gaussians. Results from all four domains could be understood within a common framework in which summation operates separately within the numerator and denominator of a contrast gain control equation. This simple arrangement of summation and counter-suppression achieves integration of various stimulus attributes without distorting the underlying contrast code.

Interactive data in financial reporting:the construction of the non-professional investor

The Securities and Exchange Commission (SEC) in the United States and in particular its immediately past chairman, Christopher Cox, has been actively promoting an upgrade of the EDGAR system of disseminating filings. The new generation of information provision has been dubbed by Chairman Cox, "Interactive Data" (SEC, 2006). In October this year the Office of Interactive Disclosure was created(http://www.sec.gov/news/press/2007/2007-213.htm). The focus of this paper is to examine the way in which the non-professional investor has been constructed by various actors. We examine the manner in which Interactive Data has been sold as the panacea for financial market 'irregularities' by the SEC and others. The academic literature shows almost no evidence of researching non-professional investors in any real sense (Young, 2006). Both this literature and the behaviour of representatives of institutions such as the SEC and FSA appears to find it convenient to construct this class of investor in a particular form and to speak for them. We theorise the activities of the SEC and its chairman in particular over a period of about three years, both following and prior to the 'credit crunch'. Our approach is to examine a selection of the policy documents released by the SEC and other interested parties and the statements made by some of the policy makers and regulators central to the programme to advance the socio-technical project that is constituted by Interactive Data. We adopt insights from ANT and more particularly the sociology of translation (Callon, 1986; Latour, 1987, 2005; Law, 1996, 2002; Law & Singleton, 2005) to show how individuals and regulators have acted as spokespersons for this malleable class of investor. We theorise the processes of accountability to investors and others and in so doing reveal the regulatory bodies taking the regulated for granted. The possible implications of technological developments in digital reporting have been identified also by the CEO's of the six biggest audit firms in a discussion document on the role of accounting information and audit in the future of global capital markets (DiPiazza et al., 2006). The potential for digital reporting enabled through XBRL to "revolutionize the entire company reporting model" (p.16) is discussed and they conclude that the new model "should be driven by the wants of investors and other users of company information,..." (p.17; emphasis in the original). Here rather than examine the somewhat illusive and vexing question of whether adding interactive functionality to 'traditional' reports can achieve the benefits claimed for nonprofessional investors we wish to consider the rhetorical and discursive moves in which the SEC and others have engaged to present such developments as providing clearer reporting and accountability standards and serving the interests of this constructed and largely unknown group - the non-professional investor.

An introduction to forensic linguistics:language in evidence

From the accusation of plagiarism in The Da Vinci Code, to the infamous hoaxer in the Yorkshire Ripper case, the use of linguistic evidence in court and the number of linguists called to act as expert witnesses in court trials has increased rapidly in the past fifteen years. An Introduction to Forensic Linguistics: Language in Evidence provides a timely and accessible introduction to this rapidly expanding subject. Using knowledge and experience gained in legal settings – Malcolm Coulthard in his work as an expert witness and Alison Johnson in her work as a West Midlands police officer – the two authors combine an array of perspectives into a distinctly unified textbook, focusing throughout on evidence from real and often high profile cases including serial killer Harold Shipman, the Bridgewater Four and the Birmingham Six. Divided into two sections, 'The Language of the Legal Process' and 'Language as Evidence', the book covers the key topics of the field. The first section looks at legal language, the structures of legal genres and the collection and testing of evidence from the initial police interview through to examination and cross-examination in the courtroom. The second section focuses on the role of the forensic linguist, the forensic phonetician and the document examiner, as well as examining in detail the linguistic investigation of authorship and plagiarism. With research tasks, suggested reading and website references provided at the end of each chapter, An Introduction to Forensic Linguistics: Language in Evidence is the essential textbook for courses in forensic linguistics and language of the law.

Estimating conditional volatility with neural networks

It is well known that one of the obstacles to effective forecasting of exchange rates is heteroscedasticity (non-stationary conditional variance). The autoregressive conditional heteroscedastic (ARCH) model and its variants have been used to estimate a time dependent variance for many financial time series. However, such models are essentially linear in form and we can ask whether a non-linear model for variance can improve results just as non-linear models (such as neural networks) for the mean have done. In this paper we consider two neural network models for variance estimation. Mixture Density Networks (Bishop 1994, Nix and Weigend 1994) combine a Multi-Layer Perceptron (MLP) and a mixture model to estimate the conditional data density. They are trained using a maximum likelihood approach. However, it is known that maximum likelihood estimates are biased and lead to a systematic under-estimate of variance. More recently, a Bayesian approach to parameter estimation has been developed (Bishop and Qazaz 1996) that shows promise in removing the maximum likelihood bias. However, up to now, this model has not been used for time series prediction. Here we compare these algorithms with two other models to provide benchmark results: a linear model (from the ARIMA family), and a conventional neural network trained with a sum-of-squares error function (which estimates the conditional mean of the time series with a constant variance noise model). This comparison is carried out on daily exchange rate data for five currencies.

Training Bayesian networks for image segmentation

We are concerned with the problem of image segmentation in which each pixel is assigned to one of a predefined finite number of classes. In Bayesian image analysis, this requires fusing together local predictions for the class labels with a prior model of segmentations. Markov Random Fields (MRFs) have been used to incorporate some of this prior knowledge, but this not entirely satisfactory as inference in MRFs is NP-hard. The multiscale quadtree model of Bouman and Shapiro (1994) is an attractive alternative, as this is a tree-structured belief network in which inference can be carried out in linear time (Pearl 1988). It is an hierarchical model where the bottom-level nodes are pixels, and higher levels correspond to downsampled versions of the image. The conditional-probability tables (CPTs) in the belief network encode the knowledge of how the levels interact. In this paper we discuss two methods of learning the CPTs given training data, using (a) maximum likelihood and the EM algorithm and (b) emphconditional maximum likelihood (CML). Segmentations obtained using networks trained by CML show a statistically-significant improvement in performance on synthetic images. We also demonstrate the methods on a real-world outdoor-scene segmentation task.

Binocular contrast vision at and above threshold

A fundamental problem for any visual system with binocular overlap is the combination of information from the two eyes. Electrophysiology shows that binocular integration of luminance contrast occurs early in visual cortex, but a specific systems architecture has not been established for human vision. Here, we address this by performing binocular summation and monocular, binocular, and dichoptic masking experiments for horizontal 1 cycle per degree test and masking gratings. These data reject three previously published proposals, each of which predict too little binocular summation and insufficient dichoptic facilitation. However, a simple development of one of the rejected models (the twin summation model) and a completely new model (the two-stage model) provide very good fits to the data. Two features common to both models are gently accelerating (almost linear) contrast transduction prior to binocular summation and suppressive ocular interactions that contribute to contrast gain control. With all model parameters fixed, both models correctly predict (1) systematic variation in psychometric slopes, (2) dichoptic contrast matching, and (3) high levels of binocular summation for various levels of binocular pedestal contrast. A review of evidence from elsewhere leads us to favor the two-stage model. © 2006 ARVO.

A risk-based model for inspection and maintenance of cross-country petroleum pipeline

The existing method of pipeline health monitoring, which requires an entire pipeline to be inspected periodically, is both time-wasting and expensive. A risk-based model that reduces the amount of time spent on inspection has been presented. This model not only reduces the cost of maintaining petroleum pipelines, but also suggests efficient design and operation philosophy, construction methodology and logical insurance plans. The risk-based model uses Analytic Hierarchy Process (AHP), a multiple attribute decision-making technique, to identify the factors that influence failure on specific segments and analyzes their effects by determining probability of risk factors. The severity of failure is determined through consequence analysis. From this, the effect of a failure caused by each risk factor can be established in terms of cost, and the cumulative effect of failure is determined through probability analysis. The technique does not totally eliminate subjectivity, but it is an improvement over the existing inspection method.

Recognition times of different views of 56 depth-rotated objects: a note concerning Verfaillie and Boutsen (1995)

The present study examines the effect of the goodness of view on the minimal exposure time required to recognize depth-rotated objects. In a previous study, Verfaillie and Boutsen (1995) derived scales of goodness of view, using a new corpus of images of depth-rotated objects. In the present experiment, a subset of this corpus (five views of 56 objects) is used to determine the recognition exposure time for each view, by increasing exposure time across successive presentations until the object is recognized. The results indicate that, for two thirds of the objects, good views are recognized more frequently and have lower recognition exposure times than bad views.

Why are MD simulated protein folding times wrong?

The question of significant deviations of protein folding times simulated using molecular dynamics from experimental values is investigated. It is shown that in the framework of Markov State Model (MSM) describing the conformational dynamics of peptides and proteins, the folding time is very sensitive to the simulation model parameters, such as forcefield and temperature. Using two peptides as examples, we show that the deviations in the folding times can reach an order of magnitude for modest variations of the molecular model. We, therefore, conclude that the folding rate values obtained in molecular dynamics simulations have to be treated with care.

A co-ordinated business object approach for supporting tactical level management decisions

The proliferation of data throughout the strategic, tactical and operational areas within many organisations, has provided a need for the decision maker to be presented with structured information that is appropriate for achieving allocated tasks. However, despite this abundance of data, managers at all levels in the organisation commonly encounter a condition of ‘information overload’, that results in a paucity of the correct information. Specifically, this thesis will focus upon the tactical domain within the organisation and the information needs of management who reside at this level. In doing so, it will argue that the link between decision making at the tactical level in the organisation, and low-level transaction processing data, should be through a common object model that used a framework based upon knowledge leveraged from co-ordination theory. In order to achieve this, the Co-ordinated Business Object Model (CBOM) was created. Detailing a two-tier framework, the first tier models data based upon four interactive object models, namely, processes, activities, resources and actors. The second tier analyses the data captured by the four object models, and returns information that can be used to support tactical decision making. In addition, the Co-ordinated Business Object Support System (CBOSS), is a prototype tool that has been developed in order to both support the CBOM implementation, and to also demonstrate the functionality of the CBOM as a modelling approach for supporting tactical management decision making. Containing a graphical user interface, the system’s functionality allows the user to create and explore alternative implementations of an identified tactical level process. In order to validate the CBOM, three verification tests have been completed. The results provide evidence that the CBOM framework helps bridge the gap between low level transaction data, and the information that is used to support tactical level decision making.