Robust nonlinear model identification methods using forward regression

In this correspondence new robust nonlinear model construction algorithms for a large class of linear-in-the-parameters models are introduced to enhance model robustness via combined parameter regularization and new robust structural selective criteria. In parallel to parameter regularization, we use two classes of robust model selection criteria based on either experimental design criteria that optimizes model adequacy, or the predicted residual sums of squares (PRESS) statistic that optimizes model generalization capability, respectively. Three robust identification algorithms are introduced, i.e., combined A- and D-optimality with regularized orthogonal least squares algorithm, respectively; and combined PRESS statistic with regularized orthogonal least squares algorithm. A common characteristic of these algorithms is that the inherent computation efficiency associated with the orthogonalization scheme in orthogonal least squares or regularized orthogonal least squares has been extended such that the new algorithms are computationally efficient. Numerical examples are included to demonstrate effectiveness of the algorithms.

Development and experimental identification of a biomechanical model of the trunk for functional electrical stimulation control in paraplegia

Objectives. Theoretic modeling and experimental studies suggest that functional electrical stimulation (FES) can improve trunk balance in spinal cord injured subjects. This can have a positive impact on daily life, increasing the volume of bimanual workspace, improving sitting posture, and wheelchair propulsion. A closed loop controller for the stimulation is desirable, as it can potentially decrease muscle fatigue and offer better rejection to disturbances. This paper proposes a biomechanical model of the human trunk, and a procedure for its identification, to be used for the future development of FES controllers. The advantage over previous models resides in the simplicity of the solution proposed, which makes it possible to identify the model just before a stimulation session ( taking into account the variability of the muscle response to the FES). Materials and Methods. The structure of the model is based on previous research on FES and muscle physiology. Some details could not be inferred from previous studies, and were determined from experimental data. Experiments with a paraplegic volunteer were conducted in order to measure the moments exerted by the trunk-passive tissues and artificially stimulated muscles. Data for model identification and validation also were collected. Results. Using the proposed structure and identification procedure, the model could adequately reproduce the moments exerted during the experiments. The study reveals that the stimulated trunk extensors can exert maximal moment when the trunk is in the upright position. In contrast, previous studies show that able-bodied subjects can exert maximal trunk extension when flexed forward. Conclusions. The proposed model and identification procedure are a successful first step toward the development of a model-based controller for trunk FES. The model also gives information on the trunk in unique conditions, normally not observable in able-bodied subjects (ie, subject only to extensor muscles contraction).

A forward constrained selection algorithm for probabilistic neural network

A new probabilistic neural network (PNN) learning algorithm based on forward constrained selection (PNN-FCS) is proposed. An incremental learning scheme is adopted such that at each step, new neurons, one for each class, are selected from the training samples arid the weights of the neurons are estimated so as to minimize the overall misclassification error rate. In this manner, only the most significant training samples are used as the neurons. It is shown by simulation that the resultant networks of PNN-FCS have good classification performance compared to other types of classifiers, but much smaller model sizes than conventional PNN.

The importance of the relationship between scale and process in understanding long-term DOC dynamics

Concentrations of dissolved organic carbon have increased in many, but not all, surface waters across acid impacted areas of Europe and North America over the last two decades. Over the last eight years several hypotheses have been put forward to explain these increases, but none are yet accepted universally. Research in this area appears to have reached a stalemate between those favouring declining atmospheric deposition, climate change or land management as the key driver of long-term DOC trends. While it is clear that many of these factors influence DOC dynamics in soil and stream waters, their effect varies over different temporal and spatial scales. We argue that regional differences in acid deposition loading may account for the apparent discrepancies between studies. DOC has shown strong monotonic increases in areas which have experienced strong downward trends in pollutant sulphur and/or seasalt deposition. Elsewhere climatic factors, that strongly influence seasonality, have also dominated inter-annual variability, and here long-term monotonic DOC trends are often difficult to detect. Furthermore, in areas receiving similar acid loadings, different catchment characteristics could have affected the site specific sensitivity to changes in acidity and therefore the magnitude of DOC release in response to changes in sulphur deposition. We suggest that confusion over these temporal and spatial scales of investigation has contributed unnecessarily to the disagreement over the main regional driver(s) of DOC trends, and that the data behind the majority of these studies is more compatible than is often conveyed.

Understanding of auditory discourse in older adults: the effects of syntax and working memory

Background: As people age, language-processing ability changes. While several factors modify discourse comprehension ability in older adults, syntactic complexity of auditory discourse has received scant attention. This is despite the widely researched domain of syntactic processing of single sentences in older adults. Aims: The aims of this study were to investigate the ability of healthy older adults to understand stories that differed in syntactic complexity, and its relation to working memory. Methods & Procedures: A total of 51 healthy adults (divided into three age groups) took part. They listened to brief stories (syntactically simple and syntactically complex) and had to respond to false/true comprehension probes following each story. Working memory capacity (digit span, forward and backward) was also measured. Outcomes & Results: Differences were found in the ability of healthy older adults to understand simple and complex discourse. The complex discourse in particular was more sensitive in discerning age-related language patterns. Only the complex discourse task correlated moderately with age. There was no correlation between age and simple discourse. As far as working memory is concerned, moderate correlations were found between working memory and complex discourse. Education did not correlate with discourse, neither simple, nor complex. Conclusions: Older adults may be less efficient in forming syntactically complex representations and this may be influenced by limitations in working memory.

Particle swarm optimization aided orthogonal forward regression for unified data modeling

We propose a unified data modeling approach that is equally applicable to supervised regression and classification applications, as well as to unsupervised probability density function estimation. A particle swarm optimization (PSO) aided orthogonal forward regression (OFR) algorithm based on leave-one-out (LOO) criteria is developed to construct parsimonious radial basis function (RBF) networks with tunable nodes. Each stage of the construction process determines the center vector and diagonal covariance matrix of one RBF node by minimizing the LOO statistics. For regression applications, the LOO criterion is chosen to be the LOO mean square error, while the LOO misclassification rate is adopted in two-class classification applications. By adopting the Parzen window estimate as the desired response, the unsupervised density estimation problem is transformed into a constrained regression problem. This PSO aided OFR algorithm for tunable-node RBF networks is capable of constructing very parsimonious RBF models that generalize well, and our analysis and experimental results demonstrate that the algorithm is computationally even simpler than the efficient regularization assisted orthogonal least square algorithm based on LOO criteria for selecting fixed-node RBF models. Another significant advantage of the proposed learning procedure is that it does not have learning hyperparameters that have to be tuned using costly cross validation. The effectiveness of the proposed PSO aided OFR construction procedure is illustrated using several examples taken from regression and classification, as well as density estimation applications.

Probability density estimation with tunable kernels using orthogonal forward regression

A generalized or tunable-kernel model is proposed for probability density function estimation based on an orthogonal forward regression procedure. Each stage of the density estimation process determines a tunable kernel, namely, its center vector and diagonal covariance matrix, by minimizing a leave-one-out test criterion. The kernel mixing weights of the constructed sparse density estimate are finally updated using the multiplicative nonnegative quadratic programming algorithm to ensure the nonnegative and unity constraints, and this weight-updating process additionally has the desired ability to further reduce the model size. The proposed tunable-kernel model has advantages, in terms of model generalization capability and model sparsity, over the standard fixed-kernel model that restricts kernel centers to the training data points and employs a single common kernel variance for every kernel. On the other hand, it does not optimize all the model parameters together and thus avoids the problems of high-dimensional ill-conditioned nonlinear optimization associated with the conventional finite mixture model. Several examples are included to demonstrate the ability of the proposed novel tunable-kernel model to effectively construct a very compact density estimate accurately.

Radial basis function classifier construction using particle swarm optimisation aided orthogonal forward regression

We develop a particle swarm optimisation (PSO) aided orthogonal forward regression (OFR) approach for constructing radial basis function (RBF) classifiers with tunable nodes. At each stage of the OFR construction process, the centre vector and diagonal covariance matrix of one RBF node is determined efficiently by minimising the leave-one-out (LOO) misclassification rate (MR) using a PSO algorithm. Compared with the state-of-the-art regularisation assisted orthogonal least square algorithm based on the LOO MR for selecting fixednode RBF classifiers, the proposed PSO aided OFR algorithm for constructing tunable-node RBF classifiers offers significant advantages in terms of better generalisation performance and smaller model size as well as imposes lower computational complexity in classifier construction process. Moreover, the proposed algorithm does not have any hyperparameter that requires costly tuning based on cross validation.

Early identification of stimulant treatment responders, partial responders and non-responders using objective measures in children and adolescents with hyperkinetic disorder

Background: The aim of this study was to evaluate stimulant medication response following a single dose of methylphenidate (MPH) in children and young people with hyperkinetic disorder using infrared motion analysis combined with a continuous performance task (QbTest system) as objective measures. The hypothesis was put forward that a moderate testdose of stimulant medication could determine a robust treatment response, partial response and non-response in relation to activity, attention and impulse control measures. Methods: The study included 44 children and young people between the ages of 7-18 years with a diagnosis of hyperkinetic disorder (F90 & F90.1). A single dose-protocol incorporated the time course effects of both immediate release MPH and extended release MPH (Concerta XL, Equasym XL) to determine comparable peak efficacy periods post intake. Results: A robust treatment response with objective measures reverting to the population mean was found in 37 participants (84%). Three participants (7%) demonstrated a partial response to MPH and four participants (9%) were determined as non-responders due to deteriorating activity measures together with no improvements in attention and impulse control measures. Conclusion: Objective measures provide early into prescribing the opportunity to measure treatment response and monitor adverse reactions to stimulant medication. Most treatment responders demonstrated an effective response to MPH on a moderate testdose facilitating a swift and more optimal titration process.

Assessing the accuracy of CME speed and trajectory estimates from STEREO observations through a comparison of independent methods

We have estimated the speed and direction of propagation of a number of Coronal Mass Ejections (CMEs) using single-spacecraft data from the STEREO Heliospheric Imager (HI) wide-field cameras. In general, these values are in good agreement with those predicted by Thernisien, Vourlidas, and Howard in Solar Phys. 256, 111 -aEuro parts per thousand 130 (2009) using a forward modelling method to fit CMEs imaged by the STEREO COR2 coronagraphs. The directions of the CMEs predicted by both techniques are in good agreement despite the fact that many of the CMEs under study travel in directions that cause them to fade rapidly in the HI images. The velocities estimated from both techniques are in general agreement although there are some interesting differences that may provide evidence for the influence of the ambient solar wind on the speed of CMEs. The majority of CMEs with a velocity estimated to be below 400 km s(-1) in the COR2 field of view have higher estimated velocities in the HI field of view, while, conversely, those with COR2 velocities estimated to be above 400 km s(-1) have lower estimated HI velocities. We interpret this as evidence for the deceleration of fast CMEs and the acceleration of slower CMEs by interaction with the ambient solar wind beyond the COR2 field of view. We also show that the uncertainties in our derived parameters are influenced by the range of elongations over which each CME can be tracked. In order to reduce the uncertainty in the predicted arrival time of a CME at 1 Astronomical Unit (AU) to within six hours, the CME needs to be tracked out to at least 30 degrees elongation. This is in good agreement with predictions of the accuracy of our technique based on Monte Carlo simulations.

Nonlinear model structure design and construction using orthogonal least squares and D-optimality design

A very efficient learning algorithm for model subset selection is introduced based on a new composite cost function that simultaneously optimizes the model approximation ability and model robustness and adequacy. The derived model parameters are estimated via forward orthogonal least squares, but the model subset selection cost function includes a D-optimality design criterion that maximizes the determinant of the design matrix of the subset to ensure the model robustness, adequacy, and parsimony of the final model. The proposed approach is based on the forward orthogonal least square (OLS) algorithm, such that new D-optimality-based cost function is constructed based on the orthogonalization process to gain computational advantages and hence to maintain the inherent advantage of computational efficiency associated with the conventional forward OLS approach. Illustrative examples are included to demonstrate the effectiveness of the new approach.

A mixture of experts network structure construction algorithm for modelling and control

This paper introduces a new fast, effective and practical model structure construction algorithm for a mixture of experts network system utilising only process data. The algorithm is based on a novel forward constrained regression procedure. Given a full set of the experts as potential model bases, the structure construction algorithm, formed on the forward constrained regression procedure, selects the most significant model base one by one so as to minimise the overall system approximation error at each iteration, while the gate parameters in the mixture of experts network system are accordingly adjusted so as to satisfy the convex constraints required in the derivation of the forward constrained regression procedure. The procedure continues until a proper system model is constructed that utilises some or all of the experts. A pruning algorithm of the consequent mixture of experts network system is also derived to generate an overall parsimonious construction algorithm. Numerical examples are provided to demonstrate the effectiveness of the new algorithms. The mixture of experts network framework can be applied to a wide variety of applications ranging from multiple model controller synthesis to multi-sensor data fusion.

Neurofuzzy design and model construction of nonlinear dynamical processes from data

A common problem in many data based modelling algorithms such as associative memory networks is the problem of the curse of dimensionality. In this paper, a new two-stage neurofuzzy system design and construction algorithm (NeuDeC) for nonlinear dynamical processes is introduced to effectively tackle this problem. A new simple preprocessing method is initially derived and applied to reduce the rule base, followed by a fine model detection process based on the reduced rule set by using forward orthogonal least squares model structure detection. In both stages, new A-optimality experimental design-based criteria we used. In the preprocessing stage, a lower bound of the A-optimality design criterion is derived and applied as a subset selection metric, but in the later stage, the A-optimality design criterion is incorporated into a new composite cost function that minimises model prediction error as well as penalises the model parameter variance. The utilisation of NeuDeC leads to unbiased model parameters with low parameter variance and the additional benefit of a parsimonious model structure. Numerical examples are included to demonstrate the effectiveness of this new modelling approach for high dimensional inputs.

Nonlinear model structure detection using optimum experimental design and orthogonal least squares

A very efficient learning algorithm for model subset selection is introduced based on a new composite cost function that simultaneously optimizes the model approximation ability and model adequacy. The derived model parameters are estimated via forward orthogonal least squares, but the subset selection cost function includes an A-optimality design criterion to minimize the variance of the parameter estimates that ensures the adequacy and parsimony of the final model. An illustrative example is included to demonstrate the effectiveness of the new approach.

Parameter estimation based on stacked regression and evolutionary algorithms

A new parameter-estimation algorithm, which minimises the cross-validated prediction error for linear-in-the-parameter models, is proposed, based on stacked regression and an evolutionary algorithm. It is initially shown that cross-validation is very important for prediction in linear-in-the-parameter models using a criterion called the mean dispersion error (MDE). Stacked regression, which can be regarded as a sophisticated type of cross-validation, is then introduced based on an evolutionary algorithm, to produce a new parameter-estimation algorithm, which preserves the parsimony of a concise model structure that is determined using the forward orthogonal least-squares (OLS) algorithm. The PRESS prediction errors are used for cross-validation, and the sunspot and Canadian lynx time series are used to demonstrate the new algorithms.