Description and training of neural network dynamics

Attractor properties of a popular discrete-time neural network model are illustrated through numerical simulations. The most complex dynamics is found to occur within particular ranges of parameters controlling the symmetry and magnitude of the weight matrix. A small network model is observed to produce fixed points, limit cycles, mode-locking, the Ruelle-Takens route to chaos, and the period-doubling route to chaos. Training algorithms for tuning this dynamical behaviour are discussed. Training can be an easy or difficult task, depending whether the problem requires the use of temporal information distributed over long time intervals. Such problems require training algorithms which can handle hidden nodes. The most prominent of these algorithms, back propagation through time, solves the temporal credit assignment problem in a way which can work only if the relevant information is distributed locally in time. The Moving Targets algorithm works for the more general case, but is computationally intensive, and prone to local minima.

Efficient training of RBF networks for classification.

Radial Basis Function networks with linear outputs are often used in regression problems because they can be substantially faster to train than Multi-layer Perceptrons. For classification problems, the use of linear outputs is less appropriate as the outputs are not guaranteed to represent probabilities. We show how RBFs with logistic and softmax outputs can be trained efficiently using the Fisher scoring algorithm. This approach can be used with any model which consists of a generalised linear output function applied to a model which is linear in its parameters. We compare this approach with standard non-linear optimisation algorithms on a number of datasets.

Bayesian training of mixture density networks

Mixture Density Networks (MDNs) are a well-established method for modelling the conditional probability density which is useful for complex multi-valued functions where regression methods (such as MLPs) fail. In this paper we extend earlier research of a regularisation method for a special case of MDNs to the general case using evidence based regularisation and we show how the Hessian of the MDN error function can be evaluated using R-propagation. The method is tested on two data sets and compared with early stopping.

Efficient training of RBF networks for classification

Radial Basis Function networks with linear outputs are often used in regression problems because they can be substantially faster to train than Multi-layer Perceptrons. For classification problems, the use of linear outputs is less appropriate as the outputs are not guaranteed to represent probabilities. In this paper we show how RBFs with logistic and softmax outputs can be trained efficiently using algorithms derived from Generalised Linear Models. This approach is compared with standard non-linear optimisation algorithms on a number of datasets.

Discriminative training of the hidden vector state model for semantic parsing

In this paper, we discuss how discriminative training can be applied to the hidden vector state (HVS) model in different task domains. The HVS model is a discrete hidden Markov model (HMM) in which each HMM state represents the state of a push-down automaton with a finite stack size. In previous applications, maximum-likelihood estimation (MLE) is used to derive the parameters of the HVS model. However, MLE makes a number of assumptions and unfortunately some of these assumptions do not hold. Discriminative training, without making such assumptions, can improve the performance of the HVS model by discriminating the correct hypothesis from the competing hypotheses. Experiments have been conducted in two domains: the travel domain for the semantic parsing task using the DARPA Communicator data and the Air Travel Information Services (ATIS) data and the bioinformatics domain for the information extraction task using the GENIA corpus. The results demonstrate modest improvements of the performance of the HVS model using discriminative training. In the travel domain, discriminative training of the HVS model gives a relative error reduction rate of 31 percent in F-measure when compared with MLE on the DARPA Communicator data and 9 percent on the ATIS data. In the bioinformatics domain, a relative error reduction rate of 4 percent in F-measure is achieved on the GENIA corpus.

School children's attitudes towards computers as a fucntion of gender, course subjects and availability of home computers

The attitudes of 328 British Secondary School children towards computers were examined in a cross-sectional survey. Measures of both general attitudes towards computers and affective reactions towards working with computers were examined in relation to the sex of the subject, courses studied (computer related/noncomputer related) and availability of a home computer. A differential pattern of results was observed. With respect to general attitudes towards computers, main effects were found for all three independent variables indicating that more favourable attitudes increased as a function of being male, doing computer courses and having a home computer. In contrast to this, affective reactions to working with computers was primarily related to doing computer courses, such that those doing computer courses reported more positive and less negative reactions. The practical and theoretical implications of these results are discussed.

The use of open learning in the training of neurotechnologists:a regional case study

DUE TO COPYRIGHT RESTRICTIONS ONLY AVAILABLE FOR CONSULTATION AT ASTON UNIVERSITY LIBRARY AND INFORMATION SERVICES WITH PRIOR ARRANGEMENT

The education and training of technicians in the construction industry

DUE TO COPYRIGHT RESTRICTIONS ONLY AVAILABLE FOR CONSULTATION AT ASTON UNIVERSITY LIBRARY AND INFORMATION SERVICES WITH PRIOR ARRANGEMENT

The effectiveness of an organisations' equal opportunities policy for the role and training of women, particularly in the fire service

DUE TO COPYRIGHT RESTRICTIONS ONLY AVAILABLE FOR CONSULTATION AT ASTON UNIVERSITY LIBRARY AND INFORMATION SERVICES WITH PRIOR ARRANGEMENT

Remote Development of Distance Course Using Virtual Learning Space ‘Web-Class Khpi’

An approach to a specialized website creation – club of distance courses authors – on the basis of Virtual Learning Space “Web-Class KhPI” is implemented and suggested in the article.

Special Features of Musical Education in Distant Learning (Training of Intonation Thinking)

The problem of using modern technologies in distant learning of intonation thinking is described in this article. An importance of intonation learning for musician students and the possibilities, provided by World Wide Web and multimedia technologies are the main point of this article.

Some Aspects of the Theme "Information Platform" in the Training of Students in Economic Majors

Information Platform has a major impact on the core activities and development of businesses. In this connection it is necessary for students of economics (the future leaders of such entities) to submit any problems related to the Information Platform, the risks they may pose, and exemplary approach to solve part or all of the problems and minimizing risks. The current issue examines the adaptation of the above problems when presenting them to students in economic majors. To the students are presented generalizations based on long observation of the occurrence and development of Information Platforms to the businesses in Bulgaria in a growing market economy.

Integration of Satellite and Financial Data to Model Future Economic Impact of Citrus Crops (Final Project Report)

This study analyzed the health and overall landcover of citrus crops in Florida. The analysis was completed using Landsat satellite imagery available free of charge from the University of Maryland Global Landcover Change Facility. The project hypothesized that combining citrus production (economic) data with citrus area per county derived from spectral signatures would yield correlations between observable spectral reflectance throughout the year, and the fiscal impact of citrus on local economies. A positive correlation between these two data types would allow us to predict the economic impact of citrus using spectral data analysis to determine final crop harvests.

The development of an optimized system of narcotic and explosive contraband mimics for calibration and training of biological detectors

Current commercially available mimics contain varying amounts of either the actual explosive/drug or the chemical compound of suspected interest by biological detectors. As a result, there is significant interest in determining the dominant chemical odor signatures of the mimics, often referred to as pseudos, particularly when compared to the genuine contraband material. This dissertation discusses results obtained from the analysis of drug and explosive headspace related to the odor profiles as recognized by trained detection canines. Analysis was performed through the use of headspace solid phase microextraction in conjunction with gas chromatography mass spectrometry (HS-SPME-GC-MS). Upon determination of specific odors, field trials were held using a combination of the target odors with COMPS. Piperonal was shown to be a dominant odor compound in the headspace of some ecstasy samples and a recognizable odor mimic by trained detection canines. It was also shown that detection canines could be imprinted on piperonal COMPS and correctly identify ecstasy samples at a threshold level of approximately 100ng/s. Isosafrole and/or MDP-2-POH show potential as training aid mimics for non-piperonal based MDMA. Acetic acid was shown to be dominant in the headspace of heroin samples and verified as a dominant odor in commercial vinegar samples; however, no common, secondary compound was detected in the headspace of either. Because of the similarities detected within respective explosive classes, several compounds were chosen for explosive mimics. A single based smokeless powder with a detectable level of 2,4-dinitrotoluene, a double based smokeless powder with a detectable level of nitroglycerine, 2-ethyl-1-hexanol, DMNB, ethyl centralite and diphenylamine were shown to be accurate mimics for TNT-based explosives, NG-based explosives, plastic explosives, tagged explosives, and smokeless powders, respectively. The combination of these six odors represents a comprehensive explosive odor kit with positive results for imprint on detection canines. As a proof of concept, the chemical compound PFTBA showed promise as a possible universal, non-target odor compound for comparison and calibration of detection canines and instrumentation. In a comparison study of shape versus vibration odor theory, the detection of d-methyl benzoate and methyl benzoate was explored using canine detectors. While results did not overwhelmingly substantiate either theory, shape odor theory provides a better explanation of the canine and human subject responses.