Search

Copyright of Journal of Engineering & Architectural Faculty of Eskisehir Osmangazi University / Eskişehir Osmangazi Üniversitesi Mühendislik ve Mimarlık Fakültesi Dergisi is the property of Eskisehir Osmangazi University and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Background and Objectives: In this study artificial neural network (ANN) used to predict milk test day records at 4th, 5th, 10th months of lactation duration and 305-day milk yield in Holstein dairy cows. Materials and Methods: Primary data source was consisting of 274025 milk production records of 7201 primiparuse to fourth birth Holstein cows, from two herd. Final source of data obtained from milk production records was consist of 87980 monthly milk test day records in 8798 rows which each row contains number of animal, herd, age, lactation, month of production, first to tenth monthly milk production records and 305-day milk yield. From the total of data, 50% was considered for neural network training, 20% for validation and 30% for testing. A multilayer perceptron (MLP) network with back propagation of error learning mechanism (BP) was used through different artificial neural network (ANN) structures to predict milk production. In order to optimize artificial neural network (ANN) structure three activation functions (hyperbolic tangent axon, sigmoid axon or linear hyperbolic tangent axon) and three back propagation algorithms viz. momentum, conjugate gradient (CG) and Leven-berg– Marquardt (LM) Training algorithms used in the hidden layer as well as in the output layer. Coefficient of determination, root of mean square error and mean absolute error were used to compare algorithms. Results: In prediction of milk production of 4th and 5th monthly test day records, LM algorithm with sigmoid axon activation functions and LM Training algorithm, with hyperbolic tangent Axon functions had the best performance between network structures respectively. In these net work structures R² were highest (0.725 and 0.642 respectively), RMSEs were lowest (4.785 and 5.345 respectively) and MAEs were lowest (3.715and 4.057 respectively). In prediction of 10th monthly test day milk production through three or four monthly test day records, obtained from the same lactation period, none of the structures had ability to predict milk production successfully. In prediction of 305-day milk yield, LM algorithm and hyperbolic tangent activation function had the best prediction through 3 test day records and R², RMSE and MAE as performance criteria were 0.799, 984.14 and 790.21 respectively. Also the same structure of the network had the best performance to predict 305-day milk yield through four or five initial test day records and performance criteria, Coefficient of determination, root of mean square error and mean absolute error were 0.856, 850.98 and 653.33 respectively, in ANN with four test day record as input variables and 0.904, 706.59 and 548.69 respectively, in ANN with five test day record as input variables, respectively. Conclusion: The artificial neural network designed in this study was able to predict the milk production of animals in the fourth month of lactation with a correlation coefficient of 0.84. On the other hand, the designed neural network was able to predict the total milk production of the animal in a lactation period of 305 days with appropriate accuracy. So that the correlation coefficients in using the first three, four and five monthly records of livestock for prediction were 0.89, 0.92 and 0.95 respectively. [ABSTRACT FROM AUTHOR]

A significant amount of hazardous compounds has leaked into the environment due to the widespread usage of organic dyes, and it is essential that these dangerous contaminants be removed in a sustainable way. This study used varying amounts of H2O2 (0, 0.5, 1.5, 3, and 5) mM/L to extract the dye from the aqueous solution. Furthermore, concentrations of 0.4, 1, 1.7, and 2.3 mM/L of Fe+2 as FeSO4·7H2O were also utilized. Batch Advanced Oxidation Process (AOP) was carried out under various working conditions, including: contact time (5-60 min), mixing speed (100-300 rpm), and UV light intensity (0-40 W). Utilizing experimental data, the AOP efficiency of Dispersed Red 17 Dye was calculated. Genetic Cascade-forward Neural Network (GCNN) was employed as a machine-learning tool to forecast the oxidation efficiency and the amount of dye that would be removed from the aqueous solution, specifically Dispersed Red 17. When compared to experimental data, the best model had an R² correlation value of 0.955. The findings of the importance analysis showed that the studied parameters affected the discoloration efficiency with order of: H2O2, UV, Fe+2, mixing speed, and contact time. The obtained results demonstrated the effectiveness of GCNN as a novel approach in forecasting the AOP efficiency of Dispersed Red 17 Dye. [ABSTRACT FROM AUTHOR]

Since the invention of the internet for military and academic research purposes, it has evolved to meet the demands of the increasing number of users on the network, who have their scope beyond military and academics. As the scope of the network expanded maintaining its security became a matter of increasing importance. With various users and interconnections of more diversified networks, the internet needs to be maintained as securely as possible for the transmission of sensitive information to be one hundred per cent safe; several anomalies may intrude on private networks. Several research works have been released around network security and this research seeks to add to the already existing body of knowledge by expounding on these attacks, proffering efficient measures to detect network intrusions, and introducing an ensemble classifier: a combination of 3 different machine learning algorithms. An ensemble classifier is used for detecting remote to local (R2L) attacks, which showed the lowest level of accuracy when the network dataset is tested using single machine learning models but the ensemble classifier gives an overall efficiency of 99.8%. [ABSTRACT FROM AUTHOR]

Back-Propagation neural networks, as well as RSM-DOE techniques, were used to predict the properties of various compositions of Iraqi oil, which were presented in this study. Paraffin and Aromatics' effect on petroleum properties, e.g., yield, density, calorific value, and other essential properties, were studied. The input-output data to the neural networks were obtained from existing local refineries in Iraq. Several network activation functions to simulate the hydrocracking process were tested and compared. the network function that gave satisfactory results in terms of convergence time and accuracy was adopted. The data were divided into training and testing parts. The results of the trained artificial neural network models for each one of the tested functions have been cross-validated with the experimental data. The network that compared well against this new set of data (i.e. testing data), with an average percent error always less than 3% for the various products of the hydrocracking unit were chosen for the study. Aromatics showed to have more profound effect on the Octane number at low concentrations of paraffin, while, for specific gravity and calorific value they have similar effects. As for boiling points and sulfur contents, aromatics have almost no effect at lower levels of paraffin. [ABSTRACT FROM AUTHOR]

The radiopharmaceutical 2-[fluorine-18]fluoro-2-deoxy-d-glucose (FDG) has been dominantly used in positron emission tomography (PET) scans for over 20 years, and due to its vast utility its applications have expanded and are continuing to expand into oncology, neurology, cardiology, and infectious/inflammatory diseases. More recently, the addition of artificial intelligence (AI) has enhanced nuclear medicine diagnosis and imaging with FDG-PET, and new radiopharmaceuticals such as prostate-specific membrane antigen (PSMA) and fibroblast activation protein inhibitor (FAPI) have emerged. Nuclear medicine therapy using agents such as [177Lu]-dotatate surpasses conventional treatments in terms of efficacy and side effects. This article reviews recently established evidence of FDG and non-FDG drugs and anticipates the future trajectory of nuclear medicine. [ABSTRACT FROM AUTHOR]

The study presents, mild steel plate, cut with dimensions of 60 mm x 40 mm, then welded with 100% argon gas by the TIG welding using design of experiment, Response Surface Methodology (RSM) and Artificial Neural Network optimization techniques. Welding current, gas flow rate, and voltage, have been selected as the process parameters during the TIG welding process. The effects of these process parameters on the weld strength factor were identified using analytical and computational intelligence techniques. The design of experiment, and Artificial Neural Networ optimization techniques were used to optimize the effect on Weld Strength Factor of the welded joints. An orthogonal array of the central composite design was prepared by the design of experiment (DOE) methodology in which experiments were performed duly as per this orthogonal array obtained. The 210.00A, 22.66 V, and 20.00 gas flow rate optimum setting of input parameters provides the better results for the weld strength factor. This solution was selected by design expert as the optimal solution having a desirability value of 0.880. The study reveals the successful use of artificial neural networks in predicting the weld strength for tungsten inert gas welding of mild steel plates. The mean square error was used to measure the performance of the network in each run. The mean square performance index for the network is a quadratic function. The input data are randomly divided into three sets. 70% are used to train the network, 15% are used to validate the network performance and 15% are used for the test. The validation of the network model produced a correlation value of 94.0% with a mean square error of 1.040E-4. the testing of the network model produced a correlation of 97.7% with mean square error 1.003E-5. The performance plot showed that the model developed was learning, which is expected of a very good network. The artificial network model produced predicted values for the weld strength of which the predicted values and the experimental values of the responses, closely fit and are in reasonable agreement with a high coefficient of correlation. [ABSTRACT FROM AUTHOR]

The speed at which cities are growing and developing today cannot be disregarded. Human activities and natural causes are both contributors to urban growth. The relationship between these factors is complex and the complexity makes it difficult for the human mind alone to understand cities. A model that helps reveal the complexity is needed for urban studies. Main objective of this study is to understand the effects of urban planning strategies on the future of the city by utilizing a Cellular Automata and Artificial Neural Networks based simulation model. Driving factors of urban growth according to development scenarios were used in the simulation process. Six different development scenarios were formulated according to the strategic plan of Izmir. Land use and driving factor data used in simulating scenarios were acquired from EarthExplorer and OpenStreetMap databases, and produced in QGIS. Future Land Use Simulation Model (FLUS) based on Cellular Automata (CA) and Artificial Neural Networks (ANN) was used. The results were assessed both by using FRAGSTATS which helped calculate fractal dimensions and visual analysis. Fractal dimension results of each scenario showed that the simulation model respected the overall urban complexity. A closer look at each scenario indicated the diverse local growth possibilities for different scenarios. The results show that urban simulation models when used as decision support tools promise a more inclusive and explicit planning process. [ABSTRACT FROM AUTHOR]

The beef flavor of beef tallow residue was improved by enzymatic hydrolysis followed by the Maillard reaction, and the flavor could be predicted using an artificial neural network. Five beef tallow residue hydrolysates were prepared using different enzymes. The Flavourzyme and Papain (FP) hydrolysate had low molecular weight peptides and high degree of hydrolysis and free amino acid content. We identified 49 main compounds, including aldehydes, pyrazines, and furan.Furan and pyrazine were the dominant volatile compounds in the five beef tallow residue-derived Maillard reaction products (MRPs), and their profiles and levels in the FP MRPs were high. The FP MRPs had the best sensory characteristics. The artificial neural network analysis revealed that the multiple input single output model had a better performance than the single input single output model, and the prediction accuracy was>90%, indicating that the MRPs sensory evaluation scores could be accurately predicted.

Predictive maintenance (PdM) is a type of approach for maintenance processes, allowing maintenance actions to be managed depending on the machine's current condition. Maintenance is therefore carried out before failures occur. The approach doesn’t only help avoid abrupt failures but also helps lower maintenance cost and provides possibilities to manufacturers to manage maintenance budgets in a more efficient way. A new deep neural network (DNN) architecture proposed in this study intends to bring a different approach to the predictive maintenance domain. There is an input layer in this architecture, a Long-Short term memory (LSTM) layer, a dropout layer (DO) followed by an LSTM layer, a hidden layer, and an output layer. The number of epochs used in the architecture and the batch size was determined using the Genetic Algorithm (GA). The activation function used after the output layer, DO ratio, and optimization algorithm optimizes loss function determined by using grid search (GS). This approach brings a different perspective to the literature for finding optimum parameters of LSTM. The neural network and hyperparameter optimization approach proposed in this study performs much better than existent studies regarding LSTM network usage for predictive maintenance purposes [ABSTRACT FROM AUTHOR]

Over the past decades, air transportation has expanded and big data for transportation era has emerged. Accurate travel demand information is an important issue for the transportation systems, especially for airline industry. So, “optimal seat capacity problem between origin and destination pairs” which is related to the load factor must be solved. In this study, a method for determining optimal seat capacity that can supply the highest load factor for the flight operation between any two countries has been introduced. The machine learning methods of Artificial Neural Network (ANN), Linear Regression (LR), Gradient Boosting (GB), and Random Forest (RF) have been applied and a software has been developed to solve the problem. The data set generated from The World Bank Database, which consists of thousands of features for all countries, has been used and a case study has been done for the period of 2014-2019 with Turkish Airlines. To the best of our knowledge, this is the first time that 1983 features have been used to forecast air travel demand in the literature within a model that covers all countries while previous studies cover only a few countries using far fewer features. Another valuable point of this study is the usage of the last regular data about the air transportation before COVID-19 pandemic. In other words, since many airline companies have experienced a decline in the air travel operation in 2020 due to COVID-19 pandemic, this study covers the most recent period (2014-2019) when flight operation performed on a regular basis. As a result, it has been observed that the developed model has forecasted the passenger load factor by an average error rate of 6.741% with GB, 6.763% with RF, 8.161% with ANN, and 9.619 % with LR.

Introduction: This study evaluated the brain activity based on the linear and nonlinear features of surface electroencephalography (EEG) in the Stroop Color and Word Test (SCWT) and the effect of learning in the test response and related EEG features.Materials and Methods: A total of 21 women and 19 men with physical and mental health participated in this study. Four stages of this SCWT, consistently in the first and second stages and inconsistently in the third and fourth stages, were taken twice by the participants with a 10-min interval. Besides, EEG recording was simultaneously taken for 1 minute at each stage.Results: The number of correct responses in the inconsistent stages was lower than that in the consistent stages, while the delay of correct responses was more in the consistent stages. EEG features showed that the relative power band of alpha 1 (8-10 Hz) frequency reduced during the test compared to the resting state. In contrast, the gamma 2 (40-50 Hz) frequency band showed a significant increase. There was no significant difference between various stages of the test and between two repetitions in the test indices and EEG features.Conclusion: Compared to the resting state, the relative power of alpha 1 and gamma 2 frequency bands changed during SCWT without considering the stage of the test. [ABSTRACT FROM AUTHOR]

This work highlights the use of artificial intelligence (AI) in fracture mechanics, in particular to solve complex problems such as the fracture of a corroded pipeline subjected to both internal pressure and axial compressive stress. The paper describes the use of artificial neural networks (ANN), a popular technique in AI, to replace the empirical expression of the DNV method with a system of simple equations based on weights and biases. The use of ANN avoids problems associated with theory, such as assumptions, boundary conditions and exact modeling. The choice of neural network structure was made on the basis of the required accuracy, measured by indicators such as the coefficient of determination R2 and the root mean square error MSE during the validation phase. The results obtained with this neural network model were satisfactory, showing good linear correlation with the target values and low divergence during the validation phase. The implementation of this model in computer applications facilitates the prediction of failure pressure without requiring in-depth expertise in finite element analysis.

This study aimed to determine the importance levels of mathematics lecture achievement, Turkish lecture achievement, Higher Education Admission Exam score, academic self-efficacy, attitude towards vocational education, academic motivation and mother and father education on the academic achievement of vocational schools of higher education students using the artificial neural network method. The data was obtained through 468 students from vocational schools of higher education at two different universities in Turkey. According to the quantitative research methodology, the correlational research design was used. The artificial neural network analysis results revealed that mathematics lecture achievement, Turkish lecture achievement and academic self-efficacy were the most critical variables that predicted the academic achievement of vocational schools of higher education students. These variables were followed by mother education level, father education level, attitude towards vocational education, Higher Education Admission Exam score and academic motivation. The results suggest that the effectiveness of the Higher Education Admission Exam score, which contributes very little to predict the academic achievement of vocational education students, need to be more questioned. [ABSTRACT FROM AUTHOR]

Artificial Neural Networks (ANN) are important data processing algorithms which are used for solving nonlinear problems. Through classical approaches, mathematical infrastructure and complex equations in ANN are difficult to understand. Interactive and multimedia-based courseware has the potential to overcome these difficulties. In this study, a web based educational courseware for ANN was developed to provide an effective and efficient learning environment so that the difficulties can be overcome. This interactive courseware was also enriched with animations and text-based course contents. In addition to this, the effects of ANN parameters' changes were observed directly through graphical results. In this way, users can easily understand the fundamentals and working mechanism of ANN. Without using any commercial libraries, the courseware was developed with ASP.NET, an object-oriented programming language. The courseware supports file formats such as XML, TXT, and CSV so that it can co-operate with other software. "Balance and Scale" data set was used to evaluate the performance of the courseware. 0.9918 accuracy, 1 specificity and 1 sensitivity values were achieved. When this study is compared to previous studies, improvements in terms of visuality, understandability and interactivity can clearly be identified. [ABSTRACT FROM AUTHOR]

Today, the amount of biological data types obtained are increasing every day. Among these data types are micro arrays that play an important role in cancer diagnosis. The data analysis that are carried out through traditional approaches have proven unsuccessful in delivering efficient results on data types where data complexity is high and where sampling is low. For this reason, using a hybrid algorithm by merging the effective features of two distinct algorithms will yield effective results. In this study, a classification process was performed firstly by dimension reduction on micro array data that were obtained from the tissues from patients with a tumor in their central nervous system and then by using an artificial neural network algorithm that was optimized through Fire Fly Algorithm (FF), a hybrid approach. The data obtained were compared to K Nearest Neighbors (KNN), Support Vector Machine (SVM) and Artificial Neural Networks (ANN) classification algorithms, which are frequently used in the literature. Also, the results were compared to the findings that were obtained from artificial neural networks, which are reinforced by Genetic Algorithm (GA), another hybrid approach. Then the results were shared. The performance results obtained show that hybrid approaches present a highly precise and more efficient classification process but they show a slower performance than basic classification algorithms. [ABSTRACT FROM AUTHOR]

Sales forecasting allows firms to plan their production outputs, which contributes to optimizing firms' inventory management via a cost reduction. However, not all firms have the same capacity to store all the necessary information through time. So, time-series with a short length are common within industries, and problems arise due to small time series does not fully capture sales' behavior. In this paper, we show the applicability of neural networks in a case where a company reports a short time-series given the changes in its warehouse structure. Given the neural networks independence form statistical assumptions, we use a multilayer-perceptron to get the sales forecasting of this enterprise. We find that learning rates variations do not significantly increase the computing time, and the validation fails with an error minor to five percent. [ABSTRACT FROM AUTHOR]

Successful patterning requires good control of the photolithography and etch processes. While compact litho models, mainly based on rigorous physics, can predict very well the contours printed in photoresist, pure empirical etch models are less accurate and more unstable. Compact etch models are based on geometrical kernels to compute the litho-etch biases that measure the distance between litho and etch contours. The definition of the kernels, as well as the choice of calibration patterns, is critical to get a robust etch model. This work proposes to define a set of independent and anisotropic etch kernels--"internal, external, curvature, Gaussian, z_profile"--designed to represent the finest details of the resist geometry to characterize precisely the etch bias at any point along a resist contour. By evaluating the etch kernels on various structures, it is possible to map their etch signatures in a multidimensional space and analyze them to find an optimal sampling of structures. The etch kernels evaluated on these structures were combined with experimental etch bias derived from scanning electron microscope contours to train artificial neural networks to predict etch bias. The method applied to contact and line/space layers shows an improvement in etch model prediction accuracy over standard etch model. This work emphasizes the importance of the etch kernel definition to characterize and predict complex etch effects. [ABSTRACT FROM AUTHOR]

In the last few years, supply chain management becomes more important,because of the globalization of business. By increasing complexity, levelof uncertainty and risk in the chain goes up. Hence supply chain risk managementhas become a major issue in the organization. One of the risksexisting in the supply chain is risk of suppliers. This research providesmodel for predicting supplier risk in Iran Alloy Steel Company that is thenanalyzed using Artificial Neural Networks which are capable to considernon-liner interrelations among criteria. In the model using fuzzy Delphi,seven criteria have been identified. Then by using AHP-VIKOR the risk ofsupplier calculated and the risk of suppliers were predicted. Finally, we usesensitive analysis for identification effect of every input on output

Chili oil, which contains large amounts of capsaicin and capsorubin, is one of the most consumed seasonings in China. These compounds significantly affect the quality, antioxidant activity, pungency, and color of chili oil. This study aimed to investigate the effect of stewing temperature, stewing time, and amount of oil on the capsaicin and capsorubin contents of Chinese chili oil. The partial least squares (PLS) regression and genetic algorithm-artificial neural network models were established and used to predict capsaicin and capsorubin contents. The genetic algorithm was applied to optimize the parameters of the network. The developed genetic algorithm-artificial neural network, which included ten hidden neurons, predicted capsaicin and capsorubin contents with correlation coefficients of 0.995 and 0.986, respectively. The neural network exhibited more accurate prediction and practicability compared with the PLS regression model. [ABSTRACT FROM AUTHOR]

The aim of this study was to investigate the effect of different pretreatment and reducing sugar content on furfural (F) and 5-hydroxymethylfurfural (HMF) contents of fermented lotus root by vinegar. The lotus root samples were fermented using vinegar for 15 days, at different solution concentrations and temperatures. The processing conditions were considered as inputs of neural network to predict the F and HMF contents of lotus root. Genetic algorithm was applied to optimize the structure and learning parameters of ANN. The developed genetic algorithm-artificial neural network (GA-ANN) which included 23 and 17 neurons in the first and second hidden layers, respectively, gives the lowest mean squared error (MSE). The correlation coefficient of ANN was compared with multiple linear regression-based models. The GA-ANN model was found to be a more accurate prediction method for the F and HMF contents of fermented lotus root than linear regression-based models. In addition, sensitivity analysis and Pearson's correlation coefficient were also analyzed to find out the relation between input and output variables. [ABSTRACT FROM AUTHOR]

The beef flavor of beef tallow residue was improved by enzymatic hydrolysis followed by the Maillard reaction, and the flavor could be predicted using an artificial neural network. Five beef tallow residue hydrolysates were prepared using different enzymes. The Flavourzyme and Papain (FP) hydrolysate had low molecular weight peptides and high degree of hydrolysis and free amino acid content. We identified 49 main compounds, including aldehydes, pyrazines, and furan. Furan and pyrazine were the dominant volatile compounds in the five beef tallow residue-derived Maillard reaction products (MRPs), and their profiles and levels in the FP MRPs were high. The FP MRPs had the best sensory characteristics. The artificial neural network analysis revealed that the multiple input single output model had a better performance than the single input single output model, and the prediction accuracy was>90%, indicating that the MRPs sensory evaluation scores could be accurately predicted., Competing Interests: The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (© 2022 The Authors. Published by Elsevier Ltd.)

Water quality monitoring in lakes and reservoirs using water samples and laboratorial analysis is expensive and time consuming. The use of artificial neural networks to predict water quality using satellite images shows great potential to make this process faster and at lower costs. This article discusses an indirect method to estimate the concentration of pigments (chlorophyll-a), an optically active parameter in water quality. A model based on artificial neural networks, using radial base functions architecture, was developed to predict Tucurui's Reservoir chlorophyll-a concentrations. As input to the neural networks spectral information from Landsat imagery was used, while pigment concentration were used as output information. To train and validate the model we used data from the years 1987, 1988, 1995, 1999, 2000 and 2004. The tested model showed a correlation coefficient of 0.92 for the estimation of pigment (chlorophyll-a) concentrations, indicating its applicability to predict this water quality parameter. [ABSTRACT FROM PUBLISHER]

This article presents construction project management effectiveness modelling from the construction management organization perspective. The paper reports on construction project performance data collected from construction management companies in Lithuania and the United States of America. Construction project management effectiveness model (CPMEM) was established by applying artificial neural networks (ANN) methodology. The discussions of project management effectiveness (success) factors identified in the literature were presented. Twelve key determinants factors that influence project management effectiveness in terms of construction cost variation were identified covering areas related to the project manager, project team, project planning, organization and control. The application algorithm of the CPMEM was developed. The CPMEM can be used during the competitive bidding process to evaluate management risk of a construction project and predict construction cost variation. [ABSTRACT FROM AUTHOR]

Accurate mapping of land cover at continental to global scales is currently limited by our ability to exploit the spatial, temporal, and radiometric characteristics of the available satellite data. Many ecologically and biogeochemically important landscape features are spatially extensive, but occur at scales much smaller than the ∼1-km footprint of wide-swath, polar orbiting radiometers. This is especially true for land cover changes associated with human activities. Satellite instruments that offer the appropriate spatial detail have much smaller swaths and longer repeat times, resulting in compositing intervals that are too large to resolve the time scales of these changes. In addition, the cost and effort associated with acquisition and processing of high-resolution data for large areas is often prohibitive. Methods for taking advantage of information contained in multiple-scale observations by combining data from high-resolution and moderate resolution sensors are thus of great current interest.In this paper, we retrieve land cover distributions in two different parts of the Brazilian Amazon region by estimating relationships between land cover fractions derived from 30-m resolution ETM+ and reflectance data from ∼1-km resolution MODIS and MISR. The scaling relationships are derived using a Bayesian-regularized artificial neural network (ANN) and compared to results using linear unmixing (LU). We explore the simultaneous use of two significant independent variables in terrestrial optical remote sensing, wavelength, and sun-sensor geometry, by combining nadir-adjusted MODIS reflectances in seven bands (VIS-SWIR) with multiangular (−71° to +71°) bidirectional reflectance data from MISR. This research was motivated by evidence from modeling and field studies demonstrating that: (a) the angular dependence of reflectance (e.g., from MISR) contains information about the structural composition of canopies that is complementary to the wavelength dependence; and (b) the SWIR portion of the spectrum (e.g., from MODIS) is sensitive to canopy moisture and shading conditions and, therefore, to the successional status of the ecosystem. This case study, using the Bayesian artificial neural network with combined MODIS-MISR data to estimate sub-pixel land cover fractions, yielded a quantitative improvement over spectral linear unmixing of single-angle, multispectral data. Our results suggest potential for broad-scale applicability despite a number of challenges related to tropical atmospheric conditions. [Copyright &y& Elsevier]

An improvement is presented on the simultaneous determination of two active ingredients present in unequal concentrations in injections. The analysis was carried out with spectrophotometric data and non-linear multivariate calibration methods, in particular artificial neural networks (ANNs). The presence of non-linearities caused by the major analyte concentrations which deviate from Beer's law was confirmed by plotting actual vs. predicted concentrations, and observing curvatures in the residuals for the estimated concentrations with linear methods. Mixtures of dextropropoxyphene and dipyrone have been analysed by using linear and non-linear partial least-squares (PLS and NPLSs) and ANNs. Notwithstanding the high degree of spectral overlap and the occurrence of non-linearities, rapid and simultaneous analysis has been achieved, with reasonably good accuracy and precision. A commercial sample was analysed by using the present methodology, and the obtained results show reasonably good agreement with those obtained by using high-performance liquid chromatography (HPLC) and a UV-spectrophotometric comparative methods. [ABSTRACT FROM AUTHOR]

The new genetic algorithm for training layered feedforward neural networks proposed here uses a mutation operator for performing the search behaviors of local optimization. Combining the random restart method with the local search technique, the algorithm can converge asymptocally, to the optimal solution. Test with a practical example showed that the improved genetic algorthm is more efficient than the conventional genetic algorithm. [ABSTRACT FROM AUTHOR]