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.

Convergence Results for the EM Approach to Mixtures of Experts Architectures

The Expectation-Maximization (EM) algorithm is an iterative approach to maximum likelihood parameter estimation. Jordan and Jacobs (1993) recently proposed an EM algorithm for the mixture of experts architecture of Jacobs, Jordan, Nowlan and Hinton (1991) and the hierarchical mixture of experts architecture of Jordan and Jacobs (1992). They showed empirically that the EM algorithm for these architectures yields significantly faster convergence than gradient ascent. In the current paper we provide a theoretical analysis of this algorithm. We show that the algorithm can be regarded as a variable metric algorithm with its searching direction having a positive projection on the gradient of the log likelihood. We also analyze the convergence of the algorithm and provide an explicit expression for the convergence rate. In addition, we describe an acceleration technique that yields a significant speedup in simulation experiments.

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.

Comparative eye tracking of experts and novices in web single sign-on

Security indicators in web browsers alert users to the presence of a secure connection between their computer and a web server; many studies have shown that such indicators are largely ignored by users in general. In other areas of computer security, research has shown that technical expertise can decrease user susceptibility to attacks. In this work, we examine whether computer or security expertise affects use of web browser security indicators. Our study takes place in the context of web-based single sign-on, in which a user can use credentials from a single identity provider to login to many relying websites; single sign-on is a more complex, and hence more difficult, security task for users. In our study, we used eye trackers and surveyed participants to examine the cues individuals use and those they report using, respectively. Our results show that users with security expertise are more likely to self-report looking at security indicators, and eye-tracking data shows they have longer gaze duration at security indicators than those without security expertise. However, computer expertise alone is not correlated with recorded use of security indicators. In survey questions, neither experts nor novices demonstrate a good understanding of the security consequences of web-based single sign-on.

Polyvanadate-stimulated NADH oxidation by plasma membranes-The need for a mixture of deca and meta forms of vanadate

Polyvanadate solutions obtained by extracting vanadium pentoxide with dilute alkali over a period of several hours contained increasing amounts of decavanadate as characterized by NMR and ir spectra. Those solutions having a metavanadate:decavanadate ratio in the range of 1-5 showed maximum stimulation of NADH oxidation by rat liver plasma membranes. Reduction of decavanadate, but not metavanadate, was obtained only in the presence of the plasma membrane enzyme system. High simulation of activity of NADH oxidation was obtained with a mixture of the two forms of vanadate and this further increased on lowering the pH. Addition of increasing concentrations of decavanadate to metavanadate and vice versa increased the stimulatory activity, reaching a maximum when the metavanadate:decavanadate ratio was in the range of 1-5. Increased stimulatory activity can also be obtained by reaching these ratios by conversion of decavanadate to metavanadate by alkaline phosphate degradation, and of metavanadate to decavanadate by acidification. These studies show for the first time that both deca and meta forms of vanadate present in polyvanadate solutions are needed for maximum activity of NADH oxidation.

Modeling transport through an environment crowded by a mixture of obstacles of different shapes and sizes

Many biological environments are crowded by macromolecules, organelles and cells which can impede the transport of other cells and molecules. Previous studies have sought to describe these effects using either random walk models or fractional order diffusion equations. Here we examine the transport of both a single agent and a population of agents through an environment containing obstacles of varying size and shape, whose relative densities are drawn from a specified distribution. Our simulation results for a single agent indicate that smaller obstacles are more effective at retarding transport than larger obstacles; these findings are consistent with our simulations of the collective motion of populations of agents. In an attempt to explore whether these kinds of stochastic random walk simulations can be described using a fractional order diffusion equation framework, we calibrate the solution of such a differential equation to our averaged agent density information. Our approach suggests that these kinds of commonly used differential equation models ought to be used with care since we are unable to match the solution of a fractional order diffusion equation to our data in a consistent fashion over a finite time period.

An Empirical Study of the Mixture of Time and Movements in Prices

This paper investigates the persistent pattern in the Helsinki Exchanges. The persistent pattern is analyzed using a time and a price approach. It is hypothesized that arrival times are related to movements in prices. Thus, the arrival times are defined as durations and formulated as an Autoregressive Conditional Duration (ACD) model as in Engle and Russell (1998). The prices are defined as price changes and formulated as a GARCH process including duration measures. The research question follows from market microstructure predictions about price intensities defined as time between price changes. The microstructure theory states that long transaction durations might be associated with both no news and bad news. Accordingly, short durations would be related to high volatility and long durations to low volatility. As a result, the spread will tend to be larger under intensive moments. The main findings of this study are 1) arrival times are positively autocorrelated and 2) long durations are associated with low volatility in the market.

The irradiation of 1:1 mixture of ammonia:carbon dioxide ice at 30 K using 1 kev electrons

In this Letter the results of an experimental investigation of 1 keV electron irradiation of a 1:1 ice mixture of NH3:CO2 at 30 K was made under ultrahigh vacuum (10(-9) mbar) conditions. Molecular products formed within the ice were detected and monitored using FTIR spectroscopy. The formation of ammonium ions (NH4+), cyanate ions (OCN-), CO was observed leading to the synthesis of ammonium carbamate (NH4NH2CO2). The consequences of these results for prebiotic chemistry in the interstellar medium and star forming regions are discussed. Crown Copyright (C) 2012 Published by Elsevier B.V. All rights reserved.

The irradiation of pure CH3OH and 1:1 mixture of NH3:CH3OH ices at 30 K using low energy electrons

The results of an experimental investigation of 1 keV electron irradiation of ices (deposited at 30 K) of (i) pure methanol and (ii) of a 1:1 mixture of NH3:CH3OH are reported. Molecular products formed within the ice were detected and monitored using FTIR spectroscopy. The products observed were methyl formate (H3COHCO), methane (CH4), hydroxymethyl (CH2OH), formamide (HCONH2), formic acid (HCOOH), formaldehyde (H2CO), formyl radical (HCO), cyanate ion (OCN-), isocyanic acid (HNCO), carbon monoxide (CO) and carbon dioxide (CO2). The consequences of these results for prebiotic chemistry in the interstellar medium and star forming regions are discussed. Crown Copyright (C) 2012 Published by Elsevier B. V. All rights reserved.

Determination of favorable inter-particle interactions for formation of substitutionally ordered solid phases from a binary mixture of oppositely charged colloidal suspensions

The solid phase formed by a binary mixture of oppositely charged colloidal particles can be either substitutionally ordered or substitutionally disordered depending on the nature and strength of interactions among the particles. In this work, we use Monte Carlo molecular simulations along with the Gibbs-Duhem integration technique to map out the favorable inter-particle interactions for the formation of substitutionally ordered crystalline phases from a fluid phase. The inter-particle interactions are modeled using the hard core Yukawa potential but the method can be easily extended to other systems of interest. The study obtains a map of interactions depicting regions indicating the type of the crystalline aggregate that forms upon phase transition.