Worldwide, it is forecasted that 131.5 million people will suffer from dementia by 2050, and the annual cost of care will increase from 818 billion USD in 2016 to 2 trillion USD by 2030, with burgeoning social consequences. Given a timely prediction of a dementia outcome in patients, appropriate mitigating interventions can be applied to reduce risk. However such prediction facilities need to be made available to wider populations, and these facilities cannot rely on specialised, costly and invasive testing (such as neuroimaging, cerebrospinal fluid collection, etc which constitute important instruments used in diagnosis), for interventions to have a meaningful quantitative impact. Hence an emerging need exists for the wider application of prognostic measures which can be deployed using lower cost data sources such as longitudinal records routinely collected by general practices. This paper proposes an efficient prediction modelling approach to the risk of dementia, using CPRD data collected from GP practices in UK, and based on machine learning in particular the Gradient Boosting Machines model combined with a survival model such as the Cox Proportional Hazard, encapsulated in a semi-supervised learning and model calibration methodology.

Abstract Video hashing is a novel technique of multimedia processing and finds applications in video retrieval, video copy detection, anti-piracy search and video authentication. In this paper, we propose a robust video hashing based on discrete cosine transform (DCT) and non-negative matrix decomposition (NMF). The proposed video hashing extracts secure features from a normalized video via random partition and dominant DCT coefficients, and exploits NMF to learn a compact representation from the secure features. Experiments with 2050 videos are carried out to validate efficiency of the proposed video hashing. The results show that the proposed video hashing is robust to many digital operations and reaches good discrimination. Receiver operating characteristic (ROC) curve comparisons illustrate that the proposed video hashing outperforms some state-of-the-art algorithms in classification between robustness and discrimination.

The contextualizing of large and complex data sources is crucial. Associated with analytical focus can be the addressing of bias in social media and other data sources, and associated with contextualization is Big Data calibration. European Social Survey data is used here, and the main objective is the evaluation of mental health survey data. While specific findings and outcomes are the major objectives here, relating to definition and properties of mental capital, future objectives will be as follows: to plan with metadata and ontology for further, future and rewarding integration with other data sources, both nationally and globally.

Dementia is one of the most feared illnesses that has a growing year-to-year negative global impact, having a health and social care cost higher than cancer, stroke and chronic heart disease, taken together. Without the availability of a cure, nor a standardised clinical test, the utilisation of machine learning methods to identify individuals that are at risk of developing dementia could bring a new step towards proactive intervention. This study's goal is to carry out a precursor analysis leading to building classification models with enhanced capabilities for differentiating diagnoses of CN (Cognitively Normal), MCI (Mild Cognitive Impairment) and Dementia. The predictive modelling approach we propose is based on the ReliefF method combined with statistical permutation tests for feature selection, and on model training, tuning, and testing based on algorithms such as Random Forests, Support Vector Machines, Gaussian Processes, Stochastic Gradient Boosting, and eXtreme Gradient Boosting. Stability of model performances were studied in computationally intensive Monte Carlo simulations. The results consistently show that our models accurately detect dementia, and also mild cognitive impairment patients by only using the inclusion of baseline measurements as predictors, thus illustrating the importance of baseline measurements. The best results issued from Monte Carlo were achieved by eXtreme Gradient Boosting optimised models, with an accuracy of 0.88 (SD 0.02), a sensitivity of 0.93 (SD 0.02) and a specificity of 0.94 (SD 0.01) for dementia, and a sensitivity of 0.86 (SD 0.02) and a specificity of 0.9 (SD 0.02) for mild cognitive impairment. These results support in particular future developments for a risk-based method that can identify an individual's risk of developing dementia.

Data Scientists are the masters of Big Data. Analyzing masses of versatile data leads to insights that, in turn, may connect to successful business strategies, crime prevention, or better health care just to name a few. Big Data is primarily approached as mathematical and technical challenge. This may lead to technology design that enables useful insights from Big Data. However, this technology-driven approach does not meet completely and consistently enough the variety of information consumer requirements. To catch up with the versatility of user needs, the technology aspect should probably be secondary. If we adopt a user-driven approach, we are more in the position to cope with the individual expectations and exigencies of information consumers. This article takes information discovery as the overarching paradigm in data science and explains how this perspective change may impact the view on the profession of the data scientist and, resulting from that, the curriculum for the education in data science. It reflects the result from discussions with companies participating in our student project cooperation program. These results are groundwork for the development of a curriculum framework for Applied Data Science.

We first pursue the study of how hierarchy provides a well-adapted tool for the analysis of change. Then, using time sequence-constrained hierarchical clustering, we develop the practical aspects of a new approach to wavelet regression. This provides a new way to link hierarchical relationships in a multivariate time-series data set with external signals. Violence data from the Colombian conflict in the years 1990-2004 are used throughout. We conclude with some proposals for further study on the relationship between social violence and market forces, viz., between the Colombian conflict and the U.S. narcotics market.

We study two aspects of information semantics: (i) the collection of all relationships, (ii) tracking and spotting anomaly and change. The first is implemented by endowing all relevant information spaces with a Euclidean metric in a common projected space. The second is modelled by an induced ultrametric. A very general way to achieve a Euclidean embedding of different information spaces based on cross-tabulation counts (and from other input data formats) is provided by Correspondence Analysis. From there, the induced ultrametric that we are particularly interested in takes a sequential - e.g. temporal - ordering of the data into account. We employ such a perspective to look at narrative, "the flow of thought and the flow of language" (Chafe). In application to policy decision making, we show how we can focus analysis in a small number of dimensions.

How best to quantify the information of an object, whether natural or artifact, is a problem of wide interest. A related problem is the computability of an object. We present practical examples of a new way to address this problem. By giving an appropriate representation to our objects, based on a hierarchical coding of information, we exemplify how it is remarkably easy to compute complex objects. Our algorithmic complexity is related to the length of the class of objects, rather than to the length of the object.

Representing the image to be inpainted in an appropriate sparse representation dictionary, and combining elements from Bayesian statistics and modern harmonic analysis, we introduce an expectation maximization (EM) algorithm for image inpainting and interpolation. From a statistical point of view, the inpainting/interpolation can be viewed as an estimation problem with missing data. Toward this goal, we propose the idea of using the EM mechanism in a Bayesian framework, where a sparsity promoting prior penalty is imposed on the reconstructed coefficients. The EM framework gives a principled way to establish formally the idea that missing samples can be recovered/interpolated based on sparse representations. We first introduce an easy and efficient sparse-representation-based iterative algorithm for image inpainting. Additionally, we derive its theoretical convergence properties. Compared to its competitors, this algorithm allows a high degree of flexibility to recover different structural components in the image (piecewise smooth, curvilinear, texture, etc.). We also suggest some guidelines to automatically tune the regularization parameter.

I demitted as Editor-in-Chief of The Computer Journal at the start of 2008. I have passed on the Editor-in-Chief role to an outstanding successor. Having been Editor-in-Chief for the past 7 years, I have to thank, in particular, Florence Leroy for providing so much administrative support. I would also like to thank Elaine Boyes, who heads Publishing, British Computer Society, and the team at Oxford University Press, especially Maxine Smith, Catherine Morgan and Clive Hemingway. A big word of thanks has to go to the authors and to the Journal's referees. In an ever more online world, The Computer Journal is doing very well indeed. Quality of research is the primary goal, over and above the form taken by research output. The Computer Journal continues to do great service to quality in our discipline. From 2006, pursuing a proposal from Alex Gammerman, the Computer Journal Lectures were initiated and have established themselves with great presentations by leaders of our field—Robin Milner, Tony Hoare, Alex Gammerman and Volodya Vovk, Keith van Rijsbergen, Samson Abramsky, Erol Gelenbe, and in early 2008, Steve Furber. It is with sadness that I note that someone who had intended giving a Computer Journal Lecture, Karen Spärck Jones, passed away in 2007.

The neighborhood counting measure is the number of all common neighborhoods between a pair of data points. It can be used as a similarity measure for different types of data through the notion of neighborhood: multivariate, sequence, and tree data. It has been shown that this measure is closely related to a secondary probability G, which is defined in terms of a primary probability P of interest to a problem. It has also been shown that the G probability can be estimated by aggregating neighborhood counts. The following questions can be asked: What is the relationship between this similarity measure and the primary probability P, especially for the task of classification? How does this similarity measure compare with the euclidean distance, since they are directly comparable? How does the G probability estimation compare with the popular kernel density estimation for the task of classification? These questions are answered in this paper, some theoretically and some experimentally. It is shown that G is a linear function of P and, therefore, a G-based Bayes classifier is equivalent to a P-based Bayes classifier. It is also shown that a weighted k-nearest neighbor classifier equipped with the neighborhood counting measure is, in fact, an approximation of the G-based Bayes classifier. It is further shown that the G probability leads to a probability estimator similar in spirit to the kernel density estimator. New experimental results are presented in this paper, which show that this measure compares favorably with the euclidean distance not only on multivariate data but also on time-series data. New experimental results are also presented regarding probability/density estimation. It was found that the G probability estimation can outperform the kernel density estimation in classification tasks.

This paper describes the undecimated wavelet transform and its reconstruction. In the first part, we show the relation between two well known undecimated wavelet transforms, the standard undecimated wavelet transform and the isotropic undecimated wavelet transform. Then we present new filter banks specially designed for undecimated wavelet decompositions which have some useful properties such as being robust to ringing artifacts which appear generally in wavelet-based denoising methods. A range of examples illustrates the results.

An important issue with authors is how quickly referee reports are obtained, and how quickly publication takes place following acceptance. Some distributions can be seen in Table 1. These figures have improved, in the sense of timeliness, over the past few years. Needless to say, a first set of referee comments could well be the first of a number of rounds of revision. The top download full-text articles in the 12 months from January to December 2005 are listed in Table 2. Computer Journal access statistics include the following: PDF full-text downloads in 2004 (Table 3) were 37,385; and in 2005, 55,968. Other planned articles or collections of articles include For information on the Computer Journal Lectures, see www.bcs.org/BCS/Awards/Events/cjlectures

With the semantic Web and Web services we have potential technologies to develop architectures that support the broad integration of business processes. A growing number of standards like BPEL or WSCI enable us to develop integrated business process architectures on a corporate as well as on an industrywide level. Service-oriented architectures (SOA) permit us to develop new generations of integrated business applications from reusable components and along structures that represent mundane as well as sophisticated business processes. However, experiences in business process modeling show that developing Web services and the corresponding orchestration layer is quite challenging. The coexistence of semantic Web standards and natural language could help to lower the language barrier in design and orchestration, allowing even non-tech people to be involved in application design. Translating natural language descriptions into the semantics of the respective SOA can help to facilitate and fasten the design process.

Journal Article Editorial Get access Fionn Murtagh Fionn Murtagh Department of Computer Science, Royal Holloway, University of London, Egham, Surrey TW20 0EX, UK Search for other works by this author on: Oxford Academic Google Scholar The Computer Journal, Volume 48, Issue 4, 2005, Page 381, https://doi.org/10.1093/comjnl/bxh122 Published: 01 January 2005

We survey a number of applications of the wavelet transform in time series prediction. We show how multiresolution prediction can capture short-range and long-term dependencies with only a few parameters to be estimated. We then develop a new multiresolution methodology for combined noise filtering and prediction, based on an approach which is similar to the Kalman filter. Based on considerable experimental assessment, we demonstrate the powerfulness of this methodology.

The three-dimensional wavelet transform is extremely important for image and video processing. The paper presents a number of three dimensional non-separable wavelet transforms, each of which is obtained using a lifting scheme. The performances of the new wavelets in terms of psycho-visual reconstruction quality and peak signal-to-noise ratio are compared to tensor product wavelets in a lossy image compression application.

This paper presents the WS-Talk interface layer, a concept for a structured natural language interface which virtualizes the composition of Web services at the end-user's side. As the number of available Web services grows, the representation of the context in which they are used becomes more and more an issue for both service providers and service consumers. While providers concentrate more on the technical levels of activation and communication within a service network, the users, i.e. the service consumers, want to form ad-hoc collaborations between Web services. To shield users of Web services from the technical details of finding and combining them, a semantic level has to be found that suits their specific needs and expertise. Through a semantic layer, WS Talk tries to create a level of abstraction that enables views on services expressed in natural language. To reduce the complexity of natural language, domain-specific languages are defined and their semantics are mapped onto Web services and their attributes. The domain of economics is currently used for testing architecture and usage aspects of the WS-Talk interface layer.

This paper is concerned with the multi-stream approach in speech recognition. In a given set of feature streams, there may be some features corrupted by noise. Ideally, these features should be excluded from recognition. To achieve this, a-priori knowledge about the identity, including both the number and location, of the noisy features is required. We present a method for estimating the number of noisy feature streams. This method assumes no knowledge about the noise. It is based on calculation of the reliability of each feature stream and then evaluation of the joint maximal reliability. Since this method decreases the uncertainty about the noisy features and is statistical in nature, it can also be used to increase robustness of other classification systems. We present an application of this method to model-order selection in the union models. We performed tests on the TIDIGITS database, corrupted by noises affecting various numbers of feature streams. The experimental results show that this model achieves recognition performance similar to the one obtained with a-priori knowledge about the identity of the corrupted features.

This paper describes a new deconvolution algorithm, based on both the wavelet transform and the curvelet transform. It extends previous results which were obtained for the denoising problem. Using these two different transformations in the same algorithm allows us to optimally detect in the same time isotropic features, well represented by the wavelet transform, and edges better represented by the curvelet transform. Adding a TV penalization term avoids the presence of oscillatory patterns around the edges which may appear when using multiscale methods. We illustrate the results with simulations.

We address the issues of discovering significant binary relationships in transaction datasets in a weighted setting. Traditional model of association rule mining is adapted to handle weighted association rule mining problems where each item is allowed to have a weight. The goal is to steer the mining focus to those significant relationships involving items with significant weights rather than being flooded in the combinatorial explosion of insignificant relationships. We identify the challenge of using weights in the iterative process of generating large itemsets. The problem of invalidation of the “downward closure property” in the weighted setting is solved by using an improved model of weighted support measurements and exploiting a “weighted downward closure property”. A new algorithm called WARM (Weighted Association Rule Mining) is developed based on the improved model. The algorithm is both scalable and efficient in discovering significant relationships in weighted settings as illustrated by experiments performed on simulated datasets.

We assess the use of an approximation to the Bayes factor for objectively assessing spatial segmentation models. The Bayes factor allows us to automatically determine thresholds, in multidimensional feature space, for such objectives as cloud mask definition. We compare our results with a cloud map currently provided as a data product.

We present in this paper a new method for contrast enhancement based on the curvelet transform. The curvelet transform represents edges better than wavelets, and is therefore well-suited for multiscale edge enhancement. We compare this approach with enhancement based on the wavelet transform, and the Multiscale Retinex. In a range of examples, we use edge detection and segmentation, among other processing applications, to provide for quantitative comparative evaluation. Our findings are that curvelet based enhancement out-performs other enhancement methods on noisy images, but on noiseless or near noiseless images curvelet based enhancement is not remarkably better than wavelet based enhancement.

Journal Article Statistics and Trends Get access Fionn Murtagh Fionn Murtagh Search for other works by this author on: Oxford Academic Google Scholar The Computer Journal, Volume 45, Issue 3, 2002, Page 259, https://doi.org/10.1093/comjnl/45.3.259 Published: 01 January 2002

In this paper we propose an experimental forecasting strategy taking into account the long-range dependence of aggregate network traffic, and we apply it to provide one-minute-ahead World-Wide Web (Web) traffic demand forecasts in terms of average number of bytes transferred. Recently, statistical examination of Web traces have shown evidence that Web traffic arising from file transfers exhibits a behavior that is consistent with the notion of self-similarity. Essentially, self-similarity indicates that significant burstiness is present on a wide range of time scales (i.e., the process is long-range dependent). Hence the idea of exploiting this multiscale property with a view towards discovering and capturing regularities underlying the time series which may prove useful for short-term traffic load forecasting. We carry out a wavelet transform decomposition of the original series to decompose the traffic time series into varying scales of temporal resolution, with the aim of making the underlying temporal structures more tractable. In a second step, each individual wavelet series—supposed to capture some features of the series—is fitted with a dynamical recurrent neural network (DRNN) model to output the wavelet forecast. The latter are afterwards recombined to form the next-minute Web Traffic demand. The method is applied on a large set of HTTP logs and is shown to yield good results. © 2001 John Wiley & Sons, Inc.

We use a redundant wavelet transform analysis to detect clusters in high-dimensional data spaces. We overcome Bellman's `curse of dimensionality' in such problems by (i) using some canonical ordering of observation and variable (document and term) dimensions in our data, (ii) applying a wavelet transform to such canonically ordered data, (iii) modelling the noise in wavelet space, (iv) defining significant component parts of the data as opposed to insignificant or noisy component parts, and (v) reading off the resultant clusters. The overall complexity of this innovative approach is linear in the data dimensionality. We describe a number of examples and test cases, including the clustering of high-dimensional hypertext data.

We discuss the implementation of a cartographic user interface to bibliographic and other information subspaces in astronomy. This includes a front end to two of the five premier scholarly journals in astronomy. We present a range of comparative assessments, in operational frameworks, of this approach to accessing and retrieving astronomical information. Finally, we discuss the particular role that such cartographic user interfaces can play in Web-based information seeking, and contrast this with widely used currently available search technologies.

A foundation is set forth for use of the wavelet transform as a spatial analysis tool for modelling the geographical representation of economic and financial measures used in agriculture. This provides a framework from which to estimate a smooth nonparametric function which describes complex, multivariate relationships embedded in spatial data, with the resulting maps conveying large amounts of information in a familiar format. We illustrate this approach for tasks which include the graphical presentation of information, density estimation and wavelet-based nonparametric regression. A redundant wavelet transform is used, and we detail the properties which make it particularly appropriate for these objectives.

Relevance has traditionally been linked with feature subset selection, but formalization of this link has not been attempted. In this paper, we propose two axioms for feature subset selection-sufficiency axiom and necessity axiom-based on which this link is formalized: The expected feature subset is the one which maximizes relevance. Finding the expected feature subset turns out to be NP-hard. We then devise a heuristic algorithm to find the expected subset which has a polynomial time complexity. The experimental results show that the algorithm finds good enough subset of features which, when presented to C4.5, results in better prediction accuracy.

We discuss a simple strategy aimed at improving neural network prediction accuracy, based on the combination of predictions at varying resolution levels of the domain under investigation (here: time series). First, a wavelet transform is used to decompose the time series into varying scales of temporal resolution. The latter provides a sensible decomposition of the data so that the underlying temporal structures of the original time series become more tractable. Then, a dynamical recurrent neural netork is trained on each resolution scale with the temporal-recurrent backpropagation algorithm. By virtue of its internal dynamic, this general class of dynamic connections network approximates the underlying law governing each resolution level by a system of non-linear difference equations. The individual wavelet scale forecasts are afterwards recombined to form the current estimate. The predictive ability of this strategy is assessed with the sunspot series.

Dynamical Recurrent Neural Networks (DRNN) (Aussem 1995a) are a class of fully recurrent networks obtained by modeling synapses as autoregressive filters. By virtue of their internal dynamic, these networks approximate the underlying law governing the time series by a system of nonlinear difference equations of internal variables. They therefore provide history-sensitive forecasts without having to be explicitly fed with external memory. The model is trained by a local and recursive error propagation algorithm called temporal-recurrent-backpropagation. The efficiency of the procedure benefits from the exponential decay of the gradient terms backpropagated through the adjoint network. We assess the predictive ability of the DRNN model with meterological and astronomical time series recorded around the candidate observation sites for the future VLT telescope. The hope is that reliable environmental forecasts provided with the model will allow the modern telescopes to be preset, a few hours in advance, in the most suited instrumental mode. In this perspective, the model is first appraised on precipitation measurements with traditional nonlinear AR and ARMA techniques using feedforward networks. Then we tackle a complex problem, namely the prediction of astronomical seeing, known to be a very erratic time series. A fuzzy coding approach is used to reduce the complexity of the underlying laws governing the seeing. Then, a fuzzy correspondence analysis is carried out to explore the internal relationships in the data. Based on a carefully selected set of meteorological variables at the same time-point, a nonlinear multiple regression, termed nowcasting (Murtagh et al. 1993, 1995), is carried out on the fuzzily coded seeing records. The DRNN is shown to outperform the fuzzy k-nearest neighbors method.

In the above-titled paper (ibid., vol.12, no.11, p.1088-92, Nov. 1990), parallel implementations of hierarchical clustering algorithms that achieve O(n/sup 2/) computational time complexity and thereby improve on the baseline of sequential implementations are described. The latter are stated to be O(n/sup 3/), with the exception of the single-link method. The commenter points out that state-of-the-art hierarchical clustering algorithms have O(n/sup 2/) time complexity and should be referred to in preference to the O(n/sup 3/) algorithms, which were described in many texts in the 1970s. Some further references in the parallelizing of hierarchic clustering algorithms are provided.>

A large number of non-parametric clustering algorithms from a wide range of applications in the social sciences, earth sciences, pattern recognition, and image processing, are critically appraised. These algorithms all have the common property of seeking to use a relational–usually contiguity–constraint, in addition to proximity information. The constraint is necessary in many applications for the visualisation of clustering results. The primary objective of this survey is to sketch out the major algorithmic paradigms in current use, with a view towards facilitating the task of algorithm design in this area.

It has often been asserted that since hierarchical clustering algorithms require pairwise interobject proximities, the complexity of these clustering procedures is at least O(N2). Recent work has disproved this by incorporating efficient nearest neighbour searching algorithms into the clustering algorithms. A general framework for hierarchical, agglomerative clustering algorithms is discussed here, which opens up the prospect of much improvement on current, widely-used algorithms. This ‘progress report’ details new algorithmic approaches in this area, and reviews recent results.