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 mixture of experts network parallel learning and model construction algorithms

A connection between a fuzzy neural network model with the mixture of experts network (MEN) modelling approach is established. Based on this linkage, two new neuro-fuzzy MEN construction algorithms are proposed to overcome the curse of dimensionality that is inherent in the majority of associative memory networks and/or other rule based systems. The first construction algorithm employs a function selection manager module in an MEN system. The second construction algorithm is based on a new parallel learning algorithm in which each model rule is trained independently, for which the parameter convergence property of the new learning method is established. As with the first approach, an expert selection criterion is utilised in this algorithm. These two construction methods are equivalent in their effectiveness in overcoming the curse of dimensionality by reducing the dimensionality of the regression vector, but the latter has the additional computational advantage of parallel processing. The proposed algorithms are analysed for effectiveness followed by numerical examples to illustrate their efficacy for some difficult data based modelling problems.

Hierarchical Mixtures of Experts and the EM Algorithm

We present a tree-structured architecture for supervised learning. The statistical model underlying the architecture is a hierarchical mixture model in which both the mixture coefficients and the mixture components are generalized linear models (GLIM's). Learning is treated as a maximum likelihood problem; in particular, we present an Expectation-Maximization (EM) algorithm for adjusting the parameters of the architecture. We also develop an on-line learning algorithm in which the parameters are updated incrementally. Comparative simulation results are presented in the robot dynamics domain.

Phase diagram of a model for a binary mixture of nematic molecules on a Bethe lattice

We investigate the phase diagram of a discrete version of the Maier-Saupe model with the inclusion of additional degrees of freedom to mimic a distribution of rodlike and disklike molecules. Solutions of this problem on a Bethe lattice come from the analysis of the fixed points of a set of nonlinear recursion relations. Besides the fixed points associated with isotropic and uniaxial nematic structures, there is also a fixed point associated with a biaxial nematic structure. Due to the existence of large overlaps of the stability regions, we resorted to a scheme to calculate the free energy of these structures deep in the interior of a large Cayley tree. Both thermodynamic and dynamic-stability analyses rule out the presence of a biaxial phase, in qualitative agreement with previous mean-field results.

Temperature-driven collapse of a coherent binary mixture of condensates

We present a temperature- dependent Hartree- Fock- Bogoliubov- Popov theory to analyze the properties of the equilibrium states of an homogeneous mixture of bosonic atoms in two different hyperfine states and in the presence of an internal Josephson coupling. In our calculation we show that the bistable structure of the equilibrium states at zero temperature changes when we increase the temperature of the system. We investigate two mechanisms of the disappearance of bistability. In one, near the collapse of one of the equilibrium states, the acoustical branch becomes unstable and the gap of the optical branch goes to zero. In the other, there is no divergent behavior of the system and bistability disappears at a temperature in which the two equilibrium states merge at a zero- population fraction imbalance. When we further increase the temperature, this state remains as a unique equilibrium configuration.

Management of Neuroendocrine Tumors: A Meeting of Experts from Latin America

A panel of experts from Latin America convened in Brazil, in May of 2007, for consensus recommendations regarding the management of neuroendocrine tumors ( NETs) of the gastrointestinal tract and pancreas. The recently introduced World Health Organization classification of NETs represents a step forward, but the former classification of carcinoids into foregut, midgut and hindgut is still likely to be useful in the near future. Macroscopic description of the tumor should be followed by light microscopic examination and immunohistochemical staining, whereas other techniques might not be widely available in Latin America. Surgery remains the mainstay of treatment for patients with potentially curable tumors, and adequate selection is paramount in order to optimize treatment results. Regarding systemic therapy, patients with well-differentiated tumors or islet-cell carcinomas may be categorized as having indolent disease, while patients with poorly differentiated, anaplastic, and small-cell carcinomas, or with atypical carcinoids, may be approached initially as having aggressive disease. Somatostatin analogues play a cytostatic role in indolent tumors, and chemotherapy may play a role against other, more aggressive NETs. Obviously, there is an urgent need for novel therapies that are effective against NETs. Copyright (C) 2008 S. Karger AG, Basel

Effects of a mixture of growth factors and proteins on the development of the osteogenic phenotype in human alveolar bone cell cultures

Strategies to promote bone repair have included exposure of cells to growth factor (GF) preparations from blood that generally include proteins as part of a complex mixture. This study aimed to evaluate the effects of such a mixture on different parameters of the development of the osteogenic phenotype in vitro. Osteoblastic cells were obtained by enzymatic digestion of human alveolar bone and cultured under standard osteogenic conditions until subconfluence. They were subcultured on Thermanox coverslips up to 14 days. Treated cultures were exposed during the first 7 days to osteogenic medium supplemented with a GFs + proteins mixture containing the major components found in platelet extracts [plate I et-derived growth factor-BB, transforming growth factor (TGF)-beta 1, TGF-beta 2, albumin, fibronectin, and thrombospondin] and to osteogenic medium alone thereafter. Control cultures were exposed only to the osteogenic medium. Treated cultures exhibited a significantly higher number of adherent cells from day 4 onward and of cycling cells at days 1 and 4, weak alkaline phosphatase (ALP) labeling, and significantly decreased levels of ALP activity and mRNA expression. At day 14, no Alizarin red-stained nodular areas were detected in cultures treated with GFs + proteins. Results were confirmed in the rat calvaria-derived osteogenic cell culture model. The addition of bone morphogenetic protein 7 or growth and differentiation factor 5 to treated cultures upregulated Runx2 and ALP mRNA expression, but surprisingly, ALP activity was not restored. These results showed that a mixture of GFs + proteins affects the development of the osteogenic phenotype both in human and rat cultures, leading to an increase in the number of cells, but expressed a less differentiated state.

Application of a simplified method based on regular regime approach to determine the effective moisture diffusivity of mixture of low molecular weight sugars and maltodextrin during desorption

The binary diffusivities of water in low molecular weight sugars; fructose, sucrose and a high molecular weight carbohydrate; maltodextrin (DE 11) and the effective diffusivities of water in mixtures of these sugars (sucrose, glucose, fructose) and maltodextrin (DE 11) were determined using a simplified procedure based on the Regular Regime Approach. The effective diffusivity of these mixtures exhibited both the concentration and molecular weight dependence. Surface stickiness was observed in all samples during desorption, with fructose exhibiting the highest and maltodextrin the lowest. (C) 2002 Elsevier Science Ltd. All rights reserved.

Soybean growth and yield after single tillage and species mixture of cover plants

The use of cover crops is important for the agricultural crop and soil management in order to improve the system and, consequently, to increase yield. Therefore, the present study analyzed the effect of crop residues of black oat (Avena strigosa Schreb.) (BO) and a cocktail (CO) of BO, forage turnip (Raphanus sativus L.) (FT) and common vetch (Vicia sativa L.) (V) on the emergence speed index (ESI), seedling emergence speed (SES) plant height and soybean yield in different intervals between cover crop desiccation with glyphosate 480 (3 L ha-1) and BRS 232 cultivar sowing. Plots of 5 x 2.5 m with 1 m of border received four treatments with BO cover crops and four with CO as well as a control for each cover crop, at random, with five replications. The plots were desiccated in intervals of 1, 10, 20 and 30 days before soybean seeding. The harvest was manual while yield was adjusted to 13% of moisture content. The experimental design was completely randomized with splitplots and means compared by the Scott and Knott test at 5% of significance. The results showed that CO of cover crops can be recommended for soybean to obtain a more vigorous seedling emergence, from 10 days after cover crop desiccation.

Children visual screening protocol: validation by a focus group panel of experts

Screening programs, particularly the inclusion of specific orthoptic tests to detect visual abnormalities, varies among countries. This study aims to: 1) describes expert perception of issues related with children visual screening; 2) identify specific orthoptic tests to detect visual abnormalities in children visual screening.

Can the complex mixture of sediment contaminants cause endocrine disruption on a benthic fish? A case study with Solea senegalensis

Pollution in coastal ecosystems is a serious threat to the biota and human populations there residing. Anthropogenic activities in these ecosystems are the main cause of contamination by endocrine disruption compounds (EDCs), which can interfere with hormonal regulation and cause adverse effects to growth, stress response and reproduction. Although the chemical nature of many EDCs is unknown, it is believed that most are organic contaminants. Under an environmental risk assessment for a contaminated estuary (the Sado, SW Portugal), the present work intended to detect endocrine disruption in a flatsfish, Solea senegalensis Kaup, 1858, and its potential relationship to organic toxicants. Animals were collected from two areas in the estuary with distinct influences (industrial and rural) and from an external reference area. To evaluate endocrine disruption, hepatic vitellogenin (VTG) concentrations in males and gonad histology were analysed. As biomarkers of exposure to organic contaminants, cytochrome P450 (CYP1A) induction and the ethoxyresorufin-O-deethylase (EROD) activity were determined. The results were contrasted to sediment contamination levels, which are overall considered low, although the area presents a complex mixture of toxicants. Either males or females were found sexually immature and showed no significant evidence of degenerative pathologies. However, hepatic VTG concentrations in males from the industrial area in estuary were superior than those from the Reference, even reaching levels comparable to those in females, which may indicate an oestrogenic effect resulting from the complex contaminant mixture. These individuals also presented higher levels of CYP1A induction and EROD activity, which is consistent with contamination by organic substances. The combination of the results suggest that the exposure of flatfish to an environment contaminated by mixed toxicants, even at low levels, may cause endocrine disruption, therefore affecting populations, which implies the need for further research in identification of potential EDCs, their sources and risks at ecosystem scale.

Electrochromic devices incorporating biohybrid electrolytes doped with a lithium salt, an ionic liquid or a mixture of both

The sol-gel method was employed in the synthesis of di-urethane cross-linked poly( caprolactone) (PCL(530)/siloxane biohybrid ormolytes incorporating either a mixture of lithium triflate (LiCF3SO3) and the ionic liquid (IL) 1-ethyl-3-methyl imidazolium tetrafluoroborate ([Emim]BF4), or solely with [Emim]BF4 or LiCF3SO3. The ormolyte doped with [Emim]BF4 is thermally more stable and exhibits higher ionic conductivity (4 x 10-4 and 2 x 10-3 S cm-1 at 36 and 98 ºC, respectively) than those containing the LiCF3SO3/[Emim]BF4 mixture or just LiCF3SO3. The three ormolytes were employed in the production of glass/ITO/ormolyte/WO3/ITO/glass electrochromic devices (ECDs) designated as ECD@Y with Y = Li-[Emim]BF4, [Emim]BF4 and Li. The three ECDs displayed fast switching speed (ca. 30 s). ECD@Li-[Emim]BF4 exhibited an electrochromic contrast of 18.4 % and an optical density change of 0.11 in the visible region, the coloration efficiency attained at 555 nm was 159 and 80.2 cm-2 C-1 in the “on” and “off” states, respectively, and the open circuit memory was 48 hours. In the “on” state the CIE 1931 color space coordinates were x = 0.29 and y = 0.30, corresponding to blue color.

Mixture of bivariate Poisson regression models with an application to insurance

In a recent paper Bermúdez [2009] used bivariate Poisson regression models for ratemaking in car insurance, and included zero-inflated models to account for the excess of zeros and the overdispersion in the data set. In the present paper, we revisit this model in order to consider alternatives. We propose a 2-finite mixture of bivariate Poisson regression models to demonstrate that the overdispersion in the data requires more structure if it is to be taken into account, and that a simple zero-inflated bivariate Poisson model does not suffice. At the same time, we show that a finite mixture of bivariate Poisson regression models embraces zero-inflated bivariate Poisson regression models as a special case. Additionally, we describe a model in which the mixing proportions are dependent on covariates when modelling the way in which each individual belongs to a separate cluster. Finally, an EM algorithm is provided in order to ensure the models’ ease-of-fit. These models are applied to the same automobile insurance claims data set as used in Bermúdez [2009] and it is shown that the modelling of the data set can be improved considerably.

Multiparametric brainstem segmentation using a modified multivariate mixture of Gaussians

The human brainstem is a densely packed, complex but highly organised structure. It not only serves as a conduit for long projecting axons conveying motor and sensory information, but also is the location of multiple primary nuclei that control or modulate a vast array of functions, including homeostasis, consciousness, locomotion, and reflexive and emotive behaviours. Despite its importance, both in understanding normal brain function as well as neurodegenerative processes, it remains a sparsely studied structure in the neuroimaging literature. In part, this is due to the difficulties in imaging the internal architecture of the brainstem in vivo in a reliable and repeatable fashion. A modified multivariate mixture of Gaussians (mmMoG) was applied to the problem of multichannel tissue segmentation. By using quantitative magnetisation transfer and proton density maps acquired at 3 T with 0.8 mm isotropic resolution, tissue probability maps for four distinct tissue classes within the human brainstem were created. These were compared against an ex vivo fixated human brain, imaged at 0.5 mm, with excellent anatomical correspondence. These probability maps were used within SPM8 to create accurate individual subject segmentations, which were then used for further quantitative analysis. As an example, brainstem asymmetries were assessed across 34 right-handed individuals using voxel based morphometry (VBM) and tensor based morphometry (TBM), demonstrating highly significant differences within localised regions that corresponded to motor and vocalisation networks. This method may have important implications for future research into MRI biomarkers of pre-clinical neurodegenerative diseases such as Parkinson's disease.

Cross validation of experts versus registration methods for target localization in deep brain stimulation.

In the last five years, Deep Brain Stimulation (DBS) has become the most popular and effective surgical technique for the treatent of Parkinson's disease (PD). The Subthalamic Nucleus (STN) is the usual target involved when applying DBS. Unfortunately, the STN is in general not visible in common medical imaging modalities. Therefore, atlas-based segmentation is commonly considered to locate it in the images. In this paper, we propose a scheme that allows both, to perform a comparison between different registration algorithms and to evaluate their ability to locate the STN automatically. Using this scheme we can evaluate the expert variability against the error of the algorithms and we demonstrate that automatic STN location is possible and as accurate as the methods currently used.