Robust initialisation of Gaussian radial basis function networks using partitioned k-means clustering

Radial basis function networks can be trained quickly using linear optimisation once centres and other associated parameters have been initialised. The authors propose a small adjustment to a well accepted initialisation algorithm which improves the network accuracy over a range of problems. The algorithm is described and results are presented.

Can carbon surface oxidation shift the pore size distribution curve calculated from Ar, N(2) and CO(2) adsorption isotherms? Simulation results for a realistic carbon model

Using the virtual porous carbon model proposed by Harris et al, we study the effect of carbon surface oxidation on the pore size distribution (PSD) curve determined from simulated Ar, N(2) and CO(2) isotherms. It is assumed that surface oxidation is not destructive for the carbon skeleton, and that all pores are accessible for studied molecules (i.e., only the effect of the change of surface chemical composition is studied). The results obtained show two important things, i.e., oxidation of the carbon surface very slightly changes the absolute porosity (calculated from the geometric method of Bhattacharya and Gubbins (BG)); however, PSD curves calculated from simulated isotherms are to a greater or lesser extent affected by the presence of surface oxides. The most reliable results are obtained from Ar adsorption data. Not only is adsorption of this adsorbate practically independent from the presence of surface oxides, but, more importantly, for this molecule one can apply the slit-like model of pores as the first approach to recover the average pore diameter of a real carbon structure. For nitrogen, the effect of carbon surface chemical composition is observed due to the quadrupole moment of this molecule, and this effect shifts the PSD curves compared to Ar. The largest differences are seen for CO2, and it is clearly demonstrated that the PSD curves obtained from adsorption isotherms of this molecule contain artificial peaks and the average pore diameter is strongly influenced by the presence of electrostatic adsorbate-adsorbate as well as adsorbate-adsorbent interactions.

Simple solution for a complex problem: Proanthocyanidins, galloyl glucoses and ellagitannins fit on a single calibration curve in high performance-gel permeation chromatography

This study was undertaken to explore gel permeation chromatography (GPC) for estimating molecular weights of proanthocyanidin fractions isolated from sainfoin (Onobrychis viciifolia). The results were compared with data obtained by thiolytic degradation of the same fractions. Polystyrene, polyethylene glycol and polymethyl methacrylate standards were not suitable for estimating the molecular weights of underivatized proanthocyanidins. Therefore, a novel HPLC-GPC method was developed based on two serially connected PolarGel-L columns using DMF that contained 5% water, 1% acetic acid and 0.15 M LiBr at 0.7 ml/min and 50 degrees C. This yielded a single calibration curve for galloyl glucoses (trigalloyl glucose, pentagalloyl glucose), ellagitannins (pedunculagin, vescalagin, punicalagin, oenothein B, gemin A), proanthocyanidins (procyanidin B2, cinnamtannin B1), and several other polyphenols (catechin, epicatechin gallate, epicallocatechin gallate, amentoflavone). These GPC predicted molecular weights represented a considerable advance over previously reported HPLC-GPC methods for underivatized proanthocyanidins. (C) 2011 Elsevier B.V. All rights reserved.

Gaussian quadrature 4D-VAR

A new incremental four-dimensional variational (4D-Var) data assimilation algorithm is introduced. The algorithm does not require the computationally expensive integrations with the nonlinear model in the outer loops. Nonlinearity is accounted for by modifying the linearization trajectory of the observation operator based on integrations with the tangent linear (TL) model. This allows us to update the linearization trajectory of the observation operator in the inner loops at negligible computational cost. As a result the distinction between inner and outer loops is no longer necessary. The key idea on which the proposed 4D-Var method is based is that by using Gaussian quadrature it is possible to get an exact correspondence between the nonlinear time evolution of perturbations and the time evolution in the TL model. It is shown that J-point Gaussian quadrature can be used to derive the exact adjoint-based observation impact equations and furthermore that it is straightforward to account for the effect of multiple outer loops in these equations if the proposed 4D-Var method is used. The method is illustrated using a three-level quasi-geostrophic model and the Lorenz (1996) model.

Measures of observation impact in non-Gaussian data assimilation

ABSTRACT Non-Gaussian/non-linear data assimilation is becoming an increasingly important area of research in the Geosciences as the resolution and non-linearity of models are increased and more and more non-linear observation operators are being used. In this study, we look at the effect of relaxing the assumption of a Gaussian prior on the impact of observations within the data assimilation system. Three different measures of observation impact are studied: the sensitivity of the posterior mean to the observations, mutual information and relative entropy. The sensitivity of the posterior mean is derived analytically when the prior is modelled by a simplified Gaussian mixture and the observation errors are Gaussian. It is found that the sensitivity is a strong function of the value of the observation and proportional to the posterior variance. Similarly, relative entropy is found to be a strong function of the value of the observation. However, the errors in estimating these two measures using a Gaussian approximation to the prior can differ significantly. This hampers conclusions about the effect of the non-Gaussian prior on observation impact. Mutual information does not depend on the value of the observation and is seen to be close to its Gaussian approximation. These findings are illustrated with the particle filter applied to the Lorenz ’63 system. This article is concluded with a discussion of the appropriateness of these measures of observation impact for different situations.

Modelling and inverting complex-valued Wiener systems

We develop a complex-valued (CV) B-spline neural network approach for efficient identification and inversion of CV Wiener systems. The CV nonlinear static function in the Wiener system is represented using the tensor product of two univariate B-spline neural networks. With the aid of a least squares parameter initialisation, the Gauss-Newton algorithm effectively estimates the model parameters that include the CV linear dynamic model coefficients and B-spline neural network weights. The identification algorithm naturally incorporates the efficient De Boor algorithm with both the B-spline curve and first order derivative recursions. An accurate inverse of the CV Wiener system is then obtained, in which the inverse of the CV nonlinear static function of the Wiener system is calculated efficiently using the Gaussian-Newton algorithm based on the estimated B-spline neural network model, with the aid of the De Boor recursions. The effectiveness of our approach for identification and inversion of CV Wiener systems is demonstrated using the application of digital predistorter design for high power amplifiers with memory

Curve crossing for the reflected Levy process at zero and infinity

Let $R_{t}=\sup_{0\leq s\leq t}X_{s}-X_{t}$ be a Levy process reflected in its maximum. We give necessary and sufficient conditions for finiteness of passage times above power law boundaries at infinity. Information as to when the expected passage time for $R_{t}$ is finite, is given. We also discuss the almost sure finiteness of $\limsup_{t\to 0}R_{t}/t^{\kappa}$, for each $\kappa\geq 0$.

The Carnegie curve

The Earth’s fair weather atmospheric electric field shows, in clean air, an average daily variation which follows universal time, globally independent of the measurement position. This single diurnal cycle variation (maximum around 19UT and minimum around 03UT) is widely known as the Carnegie curve, after the geophysical survey vessel of the Carnegie Institution of Washington on which the original measurement campaigns demonstrating the universal time variation were undertaken. The Carnegie curve’s enduring importance is in providing a reference variation against which atmospheric electricity measurements are still compared; it is believed to originate from regular daily variations in atmospheric electrification associated with the different global disturbed weather regions. Details of the instrumentation, measurement principles and data obtained on the Carnegie’s seventh and final cruise are reviewed here, also deriving new harmonic coefficients allowing calculation of the Carnegie curve for different seasons. The additional harmonic analysis now identifies changes in the phasing of the maximum and minimum in the Carnegie curve, which shows a systematic seasonal variation, linked to the solstices and equinoxes, respectively.

The M curve and the performance of Spanish international new ventures

In this paper we address three challenges. First, we discuss how international new ventures (INVs) are probably not explained by the Uppsala model as there is no time for learning about foreign markets in newly born and small firms. Only in the longer term can INVs develop experiential learning to overcome the liability of foreignness as they expand abroad. Second, we advance theoretically on previous research demonstrating that the multinationality−performance relationship of INVs follows a traditional S-shaped relationship, but they first experience a ‘born global illusion’ which leads to a non-traditional M curve. Third, using a panel data analysis for the period 1994–2008 we find empirically that Spanish INVs follow an inverted U curve in the very short term, where no learning takes place, but that experience gained over time yields an M-curve relationship once learning takes place.

A Gaussian-mixture ensemble transform filter

We generalize the popular ensemble Kalman filter to an ensemble transform filter, in which the prior distribution can take the form of a Gaussian mixture or a Gaussian kernel density estimator. The design of the filter is based on a continuous formulation of the Bayesian filter analysis step. We call the new filter algorithm the ensemble Gaussian-mixture filter (EGMF). The EGMF is implemented for three simple test problems (Brownian dynamics in one dimension, Langevin dynamics in two dimensions and the three-dimensional Lorenz-63 model). It is demonstrated that the EGMF is capable of tracking systems with non-Gaussian uni- and multimodal ensemble distributions. Copyright © 2011 Royal Meteorological Society

Asymmetric output gap effects in Phillips Curve and mark-up pricing models: evidence for the US and the UK

A number of studies have found an asymmetric response of consumer price index inflation to the output gap in the US in simple Phillips curve models. We consider whether there are similar asymmetries in mark-up pricing models, that is, whether the mark-up over producers' costs also depends upon the sign of the (adjusted) output gap. The robustness of our findings to the price series is assessed, and also whether price-output responses in the UK are asymmetric.

Dependence of sea ice yield-curve shape on ice thickness

In this note, the authors discuss the contribution that frictional sliding of ice floes (or floe aggregates) past each other and pressure ridging make to the plastic yield curve of sea ice. Using results from a previous study that explicitly modeled the amount of sliding and ridging that occurs for a given global strain rate, it is noted that the relative contribution of sliding and ridging to ice stress depends upon ice thickness. The implication is that the shape and size of the plastic yield curve is dependent upon ice thickness. The yield-curve shape dependence is in addition to plastic hardening/weakening that relates the size of the yield curve to ice thickness. In most sea ice dynamics models the yield-curve shape is taken to be independent of ice thickness. The authors show that the change of the yield curve due to a change in the ice thickness can be taken into account by a weighted sum of two thickness-independent rheologies describing ridging and sliding effects separately. It would be straightforward to implement the thickness-dependent yield-curve shape described here into sea ice models used for global or regional ice prediction.