Robust fuzzy Gustafson-Kessel clustering for nonlinear system identification

This paper deals with Takagi-Sugeno (TS) fuzzy model identification of nonlinear systems using fuzzy clustering. In particular, an extended fuzzy Gustafson-Kessel (EGK) clustering algorithm, using robust competitive agglomeration (RCA), is developed for automatically constructing a TS fuzzy model from system input-output data. The EGK algorithm can automatically determine the 'optimal' number of clusters from the training data set. It is shown that the EGK approach is relatively insensitive to initialization and is less susceptible to local minima, a benefit derived from its agglomerate property. This issue is often overlooked in the current literature on nonlinear identification using conventional fuzzy clustering. Furthermore, the robust statistical concepts underlying the EGK algorithm help to alleviate the difficulty of cluster identification in the construction of a TS fuzzy model from noisy training data. A new hybrid identification strategy is then formulated, which combines the EGK algorithm with a locally weighted, least-squares method for the estimation of local sub-model parameters. The efficacy of this new approach is demonstrated through function approximation examples and also by application to the identification of an automatic voltage regulation (AVR) loop for a simulated 3 kVA laboratory micro-machine system.

The rotational broadening of V395 Carinae. Implications on the compact object's mass

Context: The masses previously obtained for the X-ray binary 2S 0921-630 inferred a compact object that was either a high-mass neutron star or low-mass black-hole, but used a previously published value for the rotational broadening (v sin i) with large uncertainties. Aims: We aim to determine an accurate mass for the compact object through an improved measurement of the secondary star's projected equatorial rotational velocity. Methods: We have used UVES echelle spectroscopy to determine the v sin i of the secondary star (V395 Car) in the low-mass X-ray binary 2S 0921-630 by comparison to an artificially broadened spectral-type template star. In addition, we have also measured v sin i from a single high signal-to-noise ratio absorption line profile calculated using the method of Least-Squares Deconvolution (LSD). Results: We determine v sin i to lie between 31.3±0.5 km s-1 to 34.7±0.5 km s-1 (assuming zero and continuum limb darkening, respectively) in disagreement with previous results based on intermediate resolution spectroscopy obtained with the 3.6 m NTT. Using our revised v sin i value in combination with the secondary star's radial velocity gives a binary mass ratio of 0.281±0.034. Furthermore, assuming a binary inclination angle of 75° gives a compact object mass of 1.37±0.13 M_?. Conclusions: We find that using relatively low-resolution spectroscopy can result in systemic uncertainties in the measured v sin i values obtained using standard methods. We suggest the use of LSD as a secondary, reliable check of the results as LSD allows one to directly discern the shape of the absorption line profile. In the light of the new v sin i measurement, we have revised down the compact object's mass, such that it is now compatible with a canonical neutron star mass.

A hierarchical multiclass support vector machine incorporated with holistic triple learning units

This paper proposes a new hierarchical learning structure, namely the holistic triple learning (HTL), for extending the binary support vector machine (SVM) to multi-classification problems. For an N-class problem, a HTL constructs a decision tree up to a depth of A leaf node of the decision tree is allowed to be placed with a holistic triple learning unit whose generalisation abilities are assessed and approved. Meanwhile, the remaining nodes in the decision tree each accommodate a standard binary SVM classifier. The holistic triple classifier is a regression model trained on three classes, whose training algorithm is originated from a recently proposed implementation technique, namely the least-squares support vector machine (LS-SVM). A major novelty with the holistic triple classifier is the reduced number of support vectors in the solution. For the resultant HTL-SVM, an upper bound of the generalisation error can be obtained. The time complexity of training the HTL-SVM is analysed, and is shown to be comparable to that of training the one-versus-one (1-vs.-1) SVM, particularly on small-scale datasets. Empirical studies show that the proposed HTL-SVM achieves competitive classification accuracy with a reduced number of support vectors compared to the popular 1-vs-1 alternative.

Robust Estimation of HDR in fMRI using H∞ Filters

Estimation and detection of the hemodynamic response (HDR) are of great importance in functional MRI (fMRI) data analysis. In this paper, we propose the use of three H 8 adaptive filters (finite memory, exponentially weighted, and time-varying) for accurate estimation and detection of the HDR. The H 8 approach is used because it safeguards against the worst case disturbances and makes no assumptions on the (statistical) nature of the signals [B. Hassibi and T. Kailath, in Proc. ICASSP, 1995, vol. 2, pp. 949-952; T. Ratnarajah and S. Puthusserypady, in Proc. 8th IEEE Workshop DSP, 1998, pp. 1483-1487]. Performances of the proposed techniques are compared to the conventional t-test method as well as the well-known LMSs and recursive least squares algorithms. Extensive numerical simulations show that the proposed methods result in better HDR estimations and activation detections.

Two-stage RBF network construction based on particle swarm optimization

The conventional radial basis function (RBF) network optimization methods, such as orthogonal least squares or the two-stage selection, can produce a sparse network with satisfactory generalization capability. However, the RBF width, as a nonlinear parameter in the network, is not easy to determine. In the aforementioned methods, the width is always pre-determined, either by trial-and-error, or generated randomly. Furthermore, all hidden nodes share the same RBF width. This will inevitably reduce the network performance, and more RBF centres may then be needed to meet a desired modelling specification. In this paper we investigate a new two-stage construction algorithm for RBF networks. It utilizes the particle swarm optimization method to search for the optimal RBF centres and their associated widths. Although the new method needs more computation than conventional approaches, it can greatly reduce the model size and improve model generalization performance. The effectiveness of the proposed technique is confirmed by two numerical simulation examples.

Fast automatic two-stage nonlinear model identification based on the extreme learning machine

It is convenient and effective to solve nonlinear problems with a model that has a linear-in-the-parameters (LITP) structure. However, the nonlinear parameters (e.g. the width of Gaussian function) of each model term needs to be pre-determined either from expert experience or through exhaustive search. An alternative approach is to optimize them by a gradient-based technique (e.g. Newton’s method). Unfortunately, all of these methods still need a lot of computations. Recently, the extreme learning machine (ELM) has shown its advantages in terms of fast learning from data, but the sparsity of the constructed model cannot be guaranteed. This paper proposes a novel algorithm for automatic construction of a nonlinear system model based on the extreme learning machine. This is achieved by effectively integrating the ELM and leave-one-out (LOO) cross validation with our two-stage stepwise construction procedure [1]. The main objective is to improve the compactness and generalization capability of the model constructed by the ELM method. Numerical analysis shows that the proposed algorithm only involves about half of the computation of orthogonal least squares (OLS) based method. Simulation examples are included to confirm the efficacy and superiority of the proposed technique.

The Oscillations of Tau Pegasi

We present extensive spectroscopic time series observations of the multiperiodic, rapidly rotating, delta Scuti star tau Pegasi. Information about the oscillations is contained within the patterns of line-profile variation of the star's blended absorption-line spectrum. We introduce the new technique of Doppler deconvolution with which to extract these patterns by modeling the intrinsic stellar spectrum and the broadening functions for each spectrum in the time series. Frequencies and modes of oscillation are identified from the variations using the technique of Fourier-Doppler imaging and a two-dimensional least-squares cleaning algorithm. We find a rich mode spectrum with degrees up to l = 20 and with frequencies below about 35 cycles day-1. Those modes with the largest amplitudes have frequencies that lie within a narrow band. We conclude that the observed spectrum can be explained if the modes of tau Peg propagate in the prograde direction with l ~= |m| and with frequencies that are about equal in the corotating frame of the star. We discuss the implications of these results for the prospect of delta Scuti seismology.

Non-radial pulsation, rotation and outburst in the Be star omega Orionis from the MuSiCoS 1998 campaign

omega Ori (HD37490, HR1934) is a Be star known to have presented variations. In order to investigate the nature and origin of its short-term and mid-term variability, a study is performed of several spectral lines (Halpha, Hdelta, HeI 4471, 4713, 4921, 5876, 6678, CII 4267, 6578, 6583, Mg II 4481, Si III 4553 and Si II 6347), based on 249 high signal-to-noise high-resolution spectra taken with 8 telescopes over 22 consecutive nights during the MuSiCoS (Multi SIte COntinuous Spectroscopy) campaign in November-December 1998. The stellar parameters are revisited and the projected rotational velocity (v sin i = 179 km s(-1)) is redetermined using several methods. With the MuSiCoS 98 dataset, a time series analysis of line-profile variations (LPVs) is performed using the Restricted Local Cleanest (RLC) algorithm and a least squares method. The behaviour of the velocity of the centroid of the lines, the equivalent widths and the apparent vsini for several lines, as well as Violet and Red components of photospheric lines affected by emission (red He i lines, Si II 6347, CII 6578, 6583) are analyzed. The non-radial pulsation (NRP) model is examined using phase diagrams and the Fourier-Doppler Imaging (FDI) method. The LPVs are consistent with a NRP mode with l = 2 or 3, \m\ = 2 with frequency 1.03 cd(-1). It is shown that an emission line outburst occurred in the middle of the campaign. Two scenarios are proposed to explain the behaviour of a dense cloud, temporarily orbiting around the star with a frequency 0.46 c d(-1), in relation to the outburst.

Regression analysis of spatial data

Many of the most interesting questions ecologists ask lead to analyses of spatial data. Yet, perhaps confused by the large number of statistical models and fitting methods available, many ecologists seem to believe this is best left to specialists. Here, we describe the issues that need consideration when analysing spatial data and illustrate these using simulation studies. Our comparative analysis involves using methods including generalized least squares, spatial filters, wavelet revised models, conditional autoregressive models and generalized additive mixed models to estimate regression coefficients from synthetic but realistic data sets, including some which violate standard regression assumptions. We assess the performance of each method using two measures and using statistical error rates for model selection. Methods that performed well included generalized least squares family of models and a Bayesian implementation of the conditional auto-regressive model. Ordinary least squares also performed adequately in the absence of model selection, but had poorly controlled Type I error rates and so did not show the improvements in performance under model selection when using the above methods. Removing large-scale spatial trends in the response led to poor performance. These are empirical results; hence extrapolation of these findings to other situations should be performed cautiously. Nevertheless, our simulation-based approach provides much stronger evidence for comparative analysis than assessments based on single or small numbers of data sets, and should be considered a necessary foundation for statements of this type in future.

Prenatal and postnatal inflammation in relation to cortisol levels in preterm infants at 18 months corrected age

Objective

To examine whether early inflammation is related to cortisol levels at 18 months corrected age (CA) in children born very preterm.

Study Design

Infants born ≤ 32 weeks gestational age were recruited in the NICU, and placental histopathology, MRI, and chart review were obtained. At 18 months CA developmental assessment and collection of 3 salivary cortisol samples were carried out. Generalized least squares was used to analyze data from 85 infants providing 222 cortisol samples.

Results

Infants exposed to chorioamnionitis with funisitis had a significantly different pattern of cortisol across the samples compared to infants with chorioamnionitis alone or no prenatal inflammation (F[4,139] = 7.3996, P <.0001). Postnatal infections, necrotizing enterocolitis and chronic lung disease were not significantly associated with the cortisol pattern at 18 months CA.

Conclusion

In children born very preterm, prenatal inflammatory stress may contribute to altered programming of the HPA axis.

Keywords: preterm, chorioamnionitis, funisitis, premature infants, hypothalamic-pituitary-adrenal axis, infection, cortisol, stress

The influence of market deregulation on fast food consumption and body mass index:A cross-national time series analysis

Objective
To investigate the effect of fast food consumption on mean population body mass index (BMI) and explore the possible influence of market deregulation on fast food consumption and BMI.

Methods
The within-country association between fast food consumption and BMI in 25 high-income member countries of the Organisation for Economic Co-operation and Development between 1999 and 2008 was explored through multivariate panel regression models, after adjustment for per capita gross domestic product, urbanization, trade openness, lifestyle indicators and other covariates. The possible mediating effect of annual per capita intake of soft drinks, animal fats and total calories on the association between fast food consumption and BMI was also analysed. Two-stage least squares regression models were conducted, using economic freedom as an instrumental variable, to study the causal effect of fast food consumption on BMI.

Findings
After adjustment for covariates, each 1-unit increase in annual fast food transactions per capita was associated with an increase of 0.033 kg/m2 in age-standardized BMI (95% confidence interval, CI: 0.013–0.052). Only the intake of soft drinks – not animal fat or total calories – mediated the observed association (β: 0.030; 95% CI: 0.010–0.050). Economic freedom was an independent predictor of fast food consumption (β: 0.27; 95% CI: 0.16–0.37). When economic freedom was used as an instrumental variable, the association between fast food and BMI weakened but remained significant (β: 0.023; 95% CI: 0.001–0.045).

Conclusion
Fast food consumption is an independent predictor of mean BMI in high-income countries. Market deregulation policies may contribute to the obesity epidemic by facilitating the spread of fast food.

Difference equation approach to two-thermocouple sensor characterization in constant velocity flow environments

Thermocouples are one of the most popular devices for temperature measurement due to their robustness, ease of manufacture and installation, and low cost. However, when used in certain harsh environments, for example, in combustion systems and engine exhausts, large wire diameters are required, and consequently the measurement bandwidth is reduced. This article discusses a software compensation technique to address the loss of high frequency fluctuations based on measurements from two thermocouples. In particular, a difference equation (DE) approach is proposed and compared with existing methods both in simulation and on experimental test rig data with constant flow velocity. It is found that the DE algorithm, combined with the use of generalized total least squares for parameter identification, provides better performance in terms of time constant estimation without any a priori assumption on the time constant ratios of the thermocouples.

Classification the geographical origin of corn distillers dried grains with solubles by near infrared reflectance spectroscopy combined with chemometrics: A feasibility study

In this study, 137 corn distillers dried grains with solubles (DDGS) samples from a range of different geographical origins (Jilin Province of China, Heilongjiang Province of China, USA and Europe) were collected and analysed. Different near infrared spectrometers combined with different chemometric packages were used in two independent laboratories to investigate the feasibility of classifying geographical origin of DDGS. Base on the same dataset, one laboratory developed a partial least square discriminant analysis model and another laboratory developed an orthogonal partial least square discriminant analysis model. Results showed that both models could perfectly classify DDGS samples from different geographical origins. These promising results encourage the development of larger scale efforts to produce datasets which can be used to differentiate the geographical origin of DDGS and such efforts are required to provide higher level food security measures on a global scale.

Estimated Glomerular Filtration Rate Decline as a Predictor of Dialysis in Kidney Transplant Recipients

BACKGROUND: It is now common for individuals to require dialysis following the failure of a kidney transplant. Management of complications and preparation for dialysis are suboptimal in this group. To aid planning, it is desirable to estimate the time to dialysis requirement. The rate of decline in the estimated glomerular filtration rate (eGFR) may be used to this end.

METHODS: This study compared the rate of eGFR decline prior to dialysis commencement between individuals with failing transplants and transplant-naïve patients. The rate of eGFR decline was also compared between transplant recipients with and without graft failure. eGFR was calculated using the four-variable MDRD equation with rate of decline calculated by least squares linear regression.

RESULTS: The annual rate of eGFR decline in incident dialysis patients with graft failure exceeded that of the transplant-naïve incident dialysis patients. In the transplant cohort, the mean annual rate of eGFR decline prior to graft failure was 7.3 ml/min/1.73 m(2) compared to 4.8 ml/min/1.73 m(2) in the transplant-naïve group (p < 0.001) and 0.35 ml/min/1.73 m(2) in recipients without graft failure (p < 0.001). Factors associated with eGFR decline were recipient age, decade of transplantation, HLA mismatch and histological evidence of chronic immunological injury.

CONCLUSIONS: Individuals with graft failure have a rapid decline in eGFR prior to dialysis commencement. To improve outcomes, dialysis planning and management of chronic kidney disease complications should be initiated earlier than in the transplant-naïve population.

An Efficient LS-SVM-Based Method for Fuzzy System Construction

This paper proposes an efficient learning mechanism to build fuzzy rule-based systems through the construction of sparse least-squares support vector machines (LS-SVMs). In addition to the significantly reduced computational complexity in model training, the resultant LS-SVM-based fuzzy system is sparser while offers satisfactory generalization capability over unseen data. It is well known that the LS-SVMs have their computational advantage over conventional SVMs in the model training process; however, the model sparseness is lost, which is the main drawback of LS-SVMs. This is an open problem for the LS-SVMs. To tackle the nonsparseness issue, a new regression alternative to the Lagrangian solution for the LS-SVM is first presented. A novel efficient learning mechanism is then proposed in this paper to extract a sparse set of support vectors for generating fuzzy IF-THEN rules. This novel mechanism works in a stepwise subset selection manner, including a forward expansion phase and a backward exclusion phase in each selection step. The implementation of the algorithm is computationally very efficient due to the introduction of a few key techniques to avoid the matrix inverse operations to accelerate the training process. The computational efficiency is also confirmed by detailed computational complexity analysis. As a result, the proposed approach is not only able to achieve the sparseness of the resultant LS-SVM-based fuzzy systems but significantly reduces the amount of computational effort in model training as well. Three experimental examples are presented to demonstrate the effectiveness and efficiency of the proposed learning mechanism and the sparseness of the obtained LS-SVM-based fuzzy systems, in comparison with other SVM-based learning techniques.