Using the EM Algorithm to Train Neural Networks: Misconceptions and a New Algorithm for Multiclass Classification

The expectation-maximization (EM) algorithm has been of considerable interest in recent years as the basis for various algorithms in application areas of neural networks such as pattern recognition. However, there exists some misconceptions concerning its application to neural networks. In this paper, we clarify these misconceptions and consider how the EM algorithm can be adopted to train multilayer perceptron (MLP) and mixture of experts (ME) networks in applications to multiclass classification. We identify some situations where the application of the EM algorithm to train MLP networks may be of limited value and discuss some ways of handling the difficulties. For ME networks, it is reported in the literature that networks trained by the EM algorithm using iteratively reweighted least squares (IRLS) algorithm in the inner loop of the M-step, often performed poorly in multiclass classification. However, we found that the convergence of the IRLS algorithm is stable and that the log likelihood is monotonic increasing when a learning rate smaller than one is adopted. Also, we propose the use of an expectation-conditional maximization (ECM) algorithm to train ME networks. Its performance is demonstrated to be superior to the IRLS algorithm on some simulated and real data sets.

Right ventricle extraction by low level and model-based algorithm

In this paper we present an algorithm as the combination of a low level morphological operation and model based Global Circular Shortest Path scheme to explore the segmentation of the Right Ventricle. Traditional morphological operations were employed to obtain the region of interest, and adjust it to generate a mask. The image cropped by the mask is then partitioned into a few overlapping regions. Global Circular Shortest Path algorithm is then applied to extract the contour from each partition. The final step is to re-assemble the partitions to create the whole contour. The technique is deemed quite reliable and robust, as this is illustrated by a very good agreement between the extracted contour and the expert manual drawing output.

The subcycled Newmark algorithm

The popular Newmark algorithm, used for implicit direct integration of structural dynamics, is extended by means of a nodal partition to permit use of different timesteps in different regions of a structural model. The algorithm developed has as a special case an explicit-explicit subcycling algorithm previously reported by Belytschko, Yen and Mullen. That algorithm has been shown, in the absence of damping or other energy dissipation, to exhibit instability over narrow timestep ranges that become narrower as the number of degrees of freedom increases, making them unlikely to be encountered in practice. The present algorithm avoids such instabilities in the case of a one to two timestep ratio (two subcycles), achieving unconditional stability in an exponential sense for a linear problem. However, with three or more subcycles, the trapezoidal rule exhibits stability that becomes conditional, falling towards that of the central difference method as the number of subcycles increases. Instabilities over narrow timestep ranges, that become narrower as the model size increases, also appear with three or more subcycles. However by moving the partition between timesteps one row of elements into the region suitable for integration with the larger timestep these the unstable timestep ranges become extremely narrow, even in simple systems with a few degrees of freedom. As well, accuracy is improved. Use of a version of the Newmark algorithm that dissipates high frequencies minimises or eliminates these narrow bands of instability. Viscous damping is also shown to remove these instabilities, at the expense of having more effect on the low frequency response.

Schistosoma japonicum cathepsin D aspartic protease cleaves human IgG and other serum components

Recombinant cathepsin D aspartic protease of Schistosoma japonicum cleaved human IgG in vitro in a time and dose-dependent manner. Optimal cleavage was seen at pH 3.6-4.5; modest cleavage remained at pH 5.0, and no cleavage was detected above pH 5.0. Amino terminal sequencing of the major cleavage fragments of human IgG identified a Fab fragment from the VH1 domain, and 2 cleavage sites in the CH2 domain below the hinge region. The P1 and P1' residues at the 2 CH2 cleavage sites were Phe254-Leu255 and Leu325-Thr326, indicating a preference by the schistosome protease for bulky hydrophobic residues flanking the scissile bond. No cleavage of the immunoglobulin light chain was detected. In addition, the recombinant schistosome protease indiscriminately degraded the human serum proteins complement C3 and serum albumin into numerous small fragments. These results demonstrate specific cleavage of human IgG by the recombinant schistosome aspartic protease, and highlight the broad range digestive specificity of the enzyme which may play a role in the degradation of host serum proteins ingested as part of the schistosome bloodmeal.

Wildlife-based tourism & increased support for nature conservation financially and otherwise: evidence from sea-turtle ecotourism at Mon Repos

The arguments of most conservationists supporting ecotourism have been based on the view that it is environmentally friendly as a resource-use and that receipts from it can counter demands to use the natural resources involved for more extractive economic purposes. But wildlife-based ecotourism can also have positive impacts in itself on the willingness of tourists to pay for wildlife conservation, strengthen the pro-conservation attitudes of tourists, and foster personal actions by them that contribute to wildlife conservation. These aspects are explored in this article on the basis of a survey of tourists visiting Mon Repos Beach near Bundaberg, Queensland, for the purpose of watching marine turtles. The results enable several of the conservation impacts of this experience on tourists to be quantified, and highlight important relationships between specific socio-economic variables and the willingness of tourists to pay for the protection of sea turtles. Furthermore, it is shown that the on-site experiences of ecotourists have positive impacts on the willingness of tourists to pay for the conservation of wildlife, and that willingness to pay is sensitive to whether or not wildlife is seen. It is suggested that in situ ecotourism is likely to be a more powerful force for fostering pro-conservation attitudes and actions among visitors than ex situ wildlife-based tourism in aquaria and zoos.

An algorithm to predict 3 ' intron splice sites in Plasmodium falciparum genomic sequences

A new algorithm, PfAGSS, for predicting 3' splice sites in Plasmodium falciparum genomic sequences is described. Application of this program to the published P. falciparum chromosome 2 and 3 data suggests that existing programs result in a high error rate in assigning 3' intron boundaries. (C) 2001 Elsevier Science B.V. All rights reserved.

GLUT4 - At the cross roads between membrane trafficking and signal transduction

GLUT4 is a mammalian facilitative glucose transporter that is highly expressed in adipose tissue and striated muscle. In response to insulin, GLUT4 moves from intracellular storage areas to the plasma membrane, thus increasing cellular glucose uptake. While the verification of this 'translocation hypothesis' (Cushman SW. Wardzala LJ. J Biol Chem 1980;255: 4758-4762 and Suzuki K, Kono T. Proc Natl Acad Sci 1980;77: 2542-2545) has increased our understanding of insulin-regulated glucose transport, a number of fundamental questions remain unanswered. Where is GLUT4 stored within the basal cell? How does GLUT4 move to the cell surface and what mechanism does insulin employ to accelerate this process) Ultimately we require a convergence of trafficking studies with research in signal transduction. However, despite more than 30 years of intensive research we have still not reached this point. The problem is complex, involving at least two separate signal transduction pathways which feed into what appears to be a very dynamic sorting process. Below we discuss some of these complexities and highlight new data that are bringing us closer to the resolution of these questions.

Parenthood and pay in Australia and the UK: evidence from workplace surveys

This article addresses the effect of parenthood on pay, examining potential reasons for-differences between Australia and the UK that are evident in spite of their similarly minimalist. male, breadwinner style approaches to work/family issues, Although cross-national differences reflect complex intersections of policy combinations, institutional frameworks, patterns of employment and gender contracts that cannot be assessed in a single analysis, the data used in this analysis uncover some of the factors that contribute to different outcomes. Motherhood penalties in the UK appear to be associated primarily with the comparatively low level of part-time earnings in that country, while higher premiums to fatherhood at least in part reflect a wider overall wage distribution. These findings reinforce the heed to interpret earnings effects of parenthood within the context of national patterns of employment and wage distribution; and highlight the breadth of strategies needed to deliver more equitable outcomes.

An approach to verifying and debugging simulation models governed by ordinary differential equations: Part 2. Residuals analysis and a case study

In Part 1 of this paper a methodology for back-to-back testing of simulation software was described. Residuals with error-dependent geometric properties were generated. A set of potential coding errors was enumerated, along with a corresponding set of feature matrices, which describe the geometric properties imposed on the residuals by each of the errors. In this part of the paper, an algorithm is developed to isolate the coding errors present by analysing the residuals. A set of errors is isolated when the subspace spanned by their combined feature matrices corresponds to that of the residuals. Individual feature matrices are compared to the residuals and classified as 'definite', 'possible' or 'impossible'. The status of 'possible' errors is resolved using a dynamic subset testing algorithm. To demonstrate and validate the testing methodology presented in Part 1 and the isolation algorithm presented in Part 2, a case study is presented using a model for biological wastewater treatment. Both single and simultaneous errors that are deliberately introduced into the simulation code are correctly detected and isolated. Copyright (C) 2003 John Wiley Sons, Ltd.

A genetic estimation algorithm for parameters of stochastic ordinary differential equations

A generic method for the estimation of parameters for Stochastic Ordinary Differential Equations (SODEs) is introduced and developed. This algorithm, called the GePERs method, utilises a genetic optimisation algorithm to minimise a stochastic objective function based on the Kolmogorov-Smirnov statistic. Numerical simulations are utilised to form the KS statistic. Further, the examination of some of the factors that improve the precision of the estimates is conducted. This method is used to estimate parameters of diffusion equations and jump-diffusion equations. It is also applied to the problem of model selection for the Queensland electricity market. (C) 2003 Elsevier B.V. All rights reserved.

Cell death and degeneration in the symbiotic dinoflagellates of the coral Stylophora pistillata during bleaching

Rising sea temperatures are increasing the incidences of mass coral bleaching (the dissociation of the coral-algal symbiosis) and coral mortality. In this study, the effects of bleaching (induced by elevated light and temperature) on the condition of symbiotic dinoflagellates (Symbiodinium sp.) within the tissue of the hard coral Stylophora pistillata (Esper) were assessed using a suite of techniques. Bleaching of S. pistillata was accompanied by declines in the maximum potential quantum yield of photosynthesis (F-v/F-m, measured using pulse amplitude modulated [PAM] fluorometry), an increase in the number of Sytox-green-stained algae (indicating compromised algal membrane integrity and cell death), an increase in 2',7'-dichlorodihydrofluroscein diacetate (H(2)DCFDA)stained algae (indicating increased oxidative stress), as well as ultrastructural changes (vacuolisation, losses of chlorophyll, and an increase in accumulation bodies). Algae expelled from S. pistillata exhibited a complete disorganisation of cellular contents; expelled cells contained only amorphous material. In situ samples taken during a natural mass coral bleaching event on the Great Barrier Reef in February 2002 also revealed a high number of Sytox-labelled algae cells in symbio. Dinoflagellate degeneration during bleaching seems to be similar to the changes resulting from senescence-phase cell death in cultured algae. These data support a role for oxidative stress in the mechanism of coral bleaching and highlight the importance of algal degeneration during the bleaching of a reef coral.

Seeing music performance: Visual influences on perception and experience

Drawing from ethnographic, empirical, and historical/cultural perspectives, we examine the extent to which visual aspects of music contribute to the communication that takes place between performers and their listeners. First, we introduce a framework for understanding how media and genres shape aural and visual experiences of music. Second, we present case studies of two performances, and describe the relation between visual and aural aspects of performance. Third, we report empirical evidence that visual aspects of performance reliably influence perceptions of musical structure (pitch related features) and affective interpretations of music. Finally, we trace new and old media trajectories of aural and visual dimensions of music, and highlight how our conceptions, perceptions and appreciation of music are intertwined with technological innovation and media deployment strategies.