			Home About us Contact

Network Output (network + output)

Distribution by Scientific Domains

Earth and Environmental Science	40%

Selected Abstracts

Artificial neural networks for parameter estimation in geophysics

GEOPHYSICAL PROSPECTING, Issue 1 2000
Carlos Calderón-Macías
Artificial neural systems have been used in a variety of problems in the fields of science and engineering. Here we describe a study of the applicability of neural networks to solving some geophysical inverse problems. In particular, we study the problem of obtaining formation resistivities and layer thicknesses from vertical electrical sounding (VES) data and that of obtaining 1D velocity models from seismic waveform data. We use a two-layer feedforward neural network (FNN) that is trained to predict earth models from measured data. Part of the interest in using FNNs for geophysical inversion is that they are adaptive systems that perform a non-linear mapping between two sets of data from a given domain. In both of our applications, we train FNNs using synthetic data as input to the networks and a layer parametrization of the models as the network output. The earth models used for network training are drawn from an ensemble of random models within some prespecified parameter limits. For network training we use the back-propagation algorithm and a hybrid back-propagation,simulated-annealing method for the VES and seismic inverse problems, respectively. Other fundamental issues for obtaining accurate model parameter estimates using trained FNNs are the size of the training data, the network configuration, the description of the data and the model parametrization. Our simulations indicate that FNNs, if adequately trained, produce reasonably accurate earth models when observed data are input to the FNNs. [source]

Analysis of an unconventional cycle as a new comparison standard for practical heat engines: the circular/elliptical cycle in T,S diagram

INTERNATIONAL JOURNAL OF ENERGY RESEARCH, Issue 13 2004
Bahri Sahin
Abstract An unconventional cycle analysis in the T,S diagram has been carried out and the cycle characteristics such as thermal efficiency, work density (defined as the ratio of the network output to the maximum volume in the cycle), maximum volume and maximum pressure are determined. The obtained results for the unconventional cycle are compared with those of the Carnot cycle. It is proposed that the analysed unconventional cycle may be used as a better comparison standard than the Carnot cycle for practical heat engines when both size and thermal efficiency are considered. Copyright © 2004 John Wiley & Sons, Ltd. [source]

Accurate forecasting of the undecided population in a public opinion poll

JOURNAL OF FORECASTING, Issue 6 2002
Christopher Monterola
Abstract The problem of pollsters is addressed which is to forecast accurately the final answers of the undecided respondents to the primary question in a public opinion poll. The task is viewed as a pattern-recognition problem of correlating the answers of the respondents to the peripheral questions in the survey with their primary answers. The underlying pattern is determined with a supervised artificial neural network that is trained using the peripheral answers of the decided respondents whose primary answers are also known. With peripheral answers as inputs, the trained network outputs the most probable primary response of an undecided respondent. For a poll conducted to determine the approval rating of the (former) Philippine president, J. E. Estrada in December 1999 and March 2000, the trained network predicted with a 95% success rate the direct responses of a test population that consists of 24.57% of the decided population who were excluded in the network training set. For the undecided population (22.67% of December respondents; 23.67% of March respondents), the network predicted a final response distribution that is consistent with the approval/disapproval ratio of the decided population. Copyright © 2002 John Wiley & Sons, Ltd. [source]

Digital soil mapping using artificial neural networks

JOURNAL OF PLANT NUTRITION AND SOIL SCIENCE, Issue 1 2005
Thorsten Behrens
Abstract In the context of a growing demand of high-resolution spatial soil information for environmental planning and modeling, fast and accurate prediction methods are needed to provide high-quality digital soil maps. Thus, this study focuses on the development of a methodology based on artificial neural networks (ANN) that is able to spatially predict soil units. Within a test area in Rhineland-Palatinate (Germany), covering an area of about 600 km2, a digital soil map was predicted. Based on feed-forward ANN with the resilient backpropagation learning algorithm, the optimal network topology was determined with one hidden layer and 15 to 30 cells depending on the soil unit to be predicted. To describe the occurrence of a soil unit and to train the ANN, 69 different terrain attributes, 53 geologic-petrographic units, and 3 types of land use were extracted from existing maps and databases. 80% of the predicted soil units (n = 33) showed training errors (mean square error) of the ANN below 0.1, 43% were even below 0.05. Validation returned a mean accuracy of over 92% for the trained network outputs. Altogether, the presented methodology based on ANN and an extended digital terrain-analysis approach is time-saving and cost effective and provides remarkable results. Digitale Bodenkartierung mithilfe von Künstlichen Neuronalen Netzen Vor dem Hintergrund einer steigenden Nachfrage nach hoch auflösenden bodenkundlichen Flächeninformationen für die Umweltplanung und Modellierung werden schnelle und genaue Vorhersagemodelle benötigt, um hochqualitative Bodenprognosekarten zur Verfügung stellen zu können. Kernpunkt der hier vorgestellten Untersuchung ist daher die Entwicklung einer Methodik zur Erstellung von Bodenprognosekarten auf der Grundlage Künstlicher Neuronaler Netze (KNN). Als Untersuchungsgebiet diente eine Fläche von über 600 km2 im Pfälzer Wald. Vorwärts propagierende KNN auf Basis des "Resilent Backpropagation"-Algorithmus mit einer verdeckten Schicht aus 15 bis 30 Zellen erwiesen sich als optimal für die Prognose von Bodenformengesellschaften. Um das Auftreten einer Bodenformengesellschaft zu beschreiben und die KNN zu trainieren, wurden 69 Reliefparameter, 3 Nutzungsklassen sowie 53 geologisch-petrographische Einheiten verwendet. 80,% der vorhergesagten Bodenformengesellschaften (n = 33) zeigten Trainingsfehler (mittlerer quadratischer Fehler der KNN) von unter 0,1; 43,% sogar von unter 0,05. Die Validierung ergab Genauigkeiten in dem kartierten Gesamtraum von durchschnittlich über 92,% für die prognostizierten Bodenformengesellschaften. Zusammenfassend kann festgehalten werden, dass die vorgestellte Methodik auf der Basis von KNN und einer umfangreichen Digitalen Reliefanalyse einen zeit- und kosteneffektiven Ansatz zur Prognose von Bodenkarten darstellt, der hervorragende Ergebnisse liefern kann. [source]

Handwritten Thai Character Recognition Using Fourier Descriptors and Genetic Neural Networks

COMPUTATIONAL INTELLIGENCE, Issue 3 2002
Pisit Phokharatkul
This article presents a method to solve the rotated and scaling character recognition problem using Fourier descriptors and genetic neural networks. The contours of character image are extracted and separated between the outer contour and inner or loop contours. The loop contours are a special characteristic of Thai characters, called the head of the character. The special features of Thai characters (loop contours) are used at the rough classification stage, and Fourier descriptors with genetic neural networks are used at the fine classification stage. The Fourier descriptors detect the outer contour of a character and it is fed to network. These features are recognized by a multilayer neural network. Genetic algorithms (GAs) are utilized to help compute the weights of the neural network optimally and reduce uncertain states in the neural networks output. Experimental results have shown that the combination of the Fourier descriptors with genetic neural networks, loop features, and local curvature charateristics of similar characters are powerful tools for successfully classifying Thai characters. The recognition rate by this method is 99.12% for 1200 examples of handwritten Thai words (a total of 13,500 characters) written by 60 persons. [source]