Performance Of Multilayer Perceptron Research Articles

Determination of Important Topographic Factors for Landslide Mapping Analysis Using MLP Network

Landslide is one of the natural disasters that occur in Malaysia. Topographic factors such as elevation, slope angle, slope aspect, general curvature, plan curvature, and profile curvature are considered as the main causes of landslides. In order to determine the dominant topographic factors in landslide mapping analysis, a study was conducted and presented in this paper. There are three main stages involved in this study. The first stage is the extraction of extra topographic factors. Previous landslide studies had identified mainly six topographic factors. Seven new additional factors have been proposed in this study. They are longitude curvature, tangential curvature, cross section curvature, surface area, diagonal line length, surface roughness, and rugosity. The second stage is the specification of the weight of each factor using two methods. The methods are multilayer perceptron (MLP) network classification accuracy and Zhou's algorithm. At the third stage, the factors with higher weights were used to improve the MLP performance. Out of the thirteen factors, eight factors were considered as important factors, which are surface area, longitude curvature, diagonal length, slope angle, elevation, slope aspect, rugosity, and profile curvature. The classification accuracy of multilayer perceptron neural network has increased by 3% after the elimination of five less important factors.

Comparing performance of MLP and RBF neural network models for predicting South Africa’s energy consumption

In view of the close association between energy and economic growth, South Africa’s aspirations for higher growth, more energy is required; formulating a long-term economic development plan and implementing an energy strategy for a country /industry necessitates establishing the correct relationship between energy and the economy. As insufficient energy or a lack thereof is reported to be a major cause of social and economic poverty, it is very important to select a model to forecast the consumption of energy reasonably accurately. This study presents techniques based on the development of multilayer perceptron (MLP) and radial basis function (RBF) of artificial neural network (ANN) models, for calculating the energy consumption of South Africa’s industrial sector between 1993 and 2000. The approach examines the energy consumption in relation to the gross domestic product. The results indicate a strong agreement between model predictions and observed values, since the mean absolute percentage error is below 5%. When performance indices are compared, the RBF-based model is a more accurate predictor than the MLP model.

Artificial neural network-based harmonics extraction algorithm for shunt active power filter control

This paper presents a harmonics extraction algorithm using artificial neural network methods. The neural network algorithm was used due to the simpler calculation process compared with conventional method such as fast Fourier transform (FFT). Two types of neural network, i.e., multi-layer perceptron (MLP) and radial basis function (RBF) were employed to extract harmonics current component from its distorted wave current. Further, the extracted harmonics current was used as reference current for shunt active power filter (APF) control. This paper compared the performance of MLP and RBF for harmonics extraction. The advantages of RBF are simpler shape of the network and faster learning speed. Unfortunately, the RBF need to be trained recursively for various harmonics component. MLP can be used to extract various harmonics component in specific data range but need large number of data training hence slower training process.

MLP refined posterior features for noise robust phoneme recognition

In this research, we have investigated improvement in the accuracy and robustness of phoneme recognition by refining posterior features extracted from single stream cepstral features. The refinement process is done using Multi Layer Perceptron (MLP) in a cascaded structure. The combination of frame posterior feature vectors, along with the entropy of each frame, as a confidence measure of posterior vectors, in the context window, is used to train a refiner MLP for estimating a new phoneme posterior feature set with the advantage of more accuracy and robustness. The confidence measure, as an informative feature, would enhance the refiner MLP performance in the correction of misclassified frames. The refiner MLP also models language level phonetics and lexical knowledge, using embedded information in the phoneme posteriors of a large context window. The suitability of these refined posterior features is evaluated on the tandem connectionist structures. Results show a significant improvement in both frame classification and phoneme recognition rates on the TIMIT acoustic and phonetic corpus compared to standard posterior features.

Integrating Mahalanobis typicalities with a neural network for rubber distribution mapping

Accurate rubber distribution mapping is critical to the study of its expansion and to provide a better understanding of the consequences of land-cover and land-use change on carbon and water cycles. Employing Mahalanobis typicalities as inputs to a hard classifier to enhance the capability of generalization has not previously been explored. This letter presents a novel approach by integrating Mahalanobis typicalities with the multi-layer perceptron (MLP) neural network for mapping of rubber. A case study from the Thai–Lao and Sino–Lao borders was conducted using Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) data. Different combinations of the nine ASTER bands including Visible and Near Infrared (VNIR) and Short-wave Infrared (SWIR), Normalized Difference Vegetation Index (NDVI) and Mahalanobis typicalities were used as input variables to the MLP. Results indicate that including Mahalanobis typicalities as input variables can improve the MLP's performance and increase the user's accuracy of rubber mapping.

Modeling of Autocatalytic Hydrolysis of Adefovir Dipivoxil in Solid Formulations

The stability and hydrolysis kinetics of a phosphate prodrug, adefovir dipivoxil, in solid formulations were studied. The stability relationship between five solid formulations was explored. An autocatalytic mechanism for hydrolysis could be proposed according to the kinetic behavior which fits the Prout-Tompkins model well. For the classical kinetic models could hardly describe and predict the hydrolysis kinetics of adefovir dipivoxil in solid formulations accurately when the temperature is high, a feedforward multilayer perceptron (MLP) neural network was constructed to model the hydrolysis kinetics. The build-in approaches in Weka, such as lazy classifiers and rule-based learners (IBk, KStar, DecisionTable and M5Rules), were used to verify the performance of MLP. The predictability of the models was evaluated by 10-fold cross-validation and an external test set. It reveals that MLP should be of general applicability proposing an alternative efficient way to model and predict autocatalytic hydrolysis kinetics for phosphate prodrugs.

Performance prediction of ocean color Monte Carlo simulations using multi-layer perceptron neural networks

A performance modeling method is presented to predict the execution time of a parallel Monte Carlo (MC) radiative transfer simulation code for ocean color applications. The execution time of MC simulations is predicted using a multi-layer perceptron (MLP) neural network regression model trained with past execution time measurements in different execution environments and simulation cases. On the basis of the MLP performance model, a complementary job-environment mapping algorithm enables an efficient utilization of available high-performance computing resources minimizing the total execution time of the simulation jobs distributed in multiple environments.

Breast cancer classification applying artificial metaplasticity algorithm

A novel improvement in neural network training for pattern classification is presented in this paper. The proposed training algorithm is inspired by the biological metaplasticity property of neurons and Shannon's information theory. This algorithm is applicable to artificial neural networks (ANNs) in general, although here it is applied to a multilayer perceptron (MLP). During the training phase, the artificial metaplasticity multilayer perceptron (AMMLP) algorithm assigns higher values for updating the weights in the less frequent activations than in the more frequent ones. AMMLP achieves a more efficient training and improves MLP performance. The well-known and readily available Wisconsin Breast Cancer Database (WBCD) has been used to test the algorithm. Performance of the AMMLP was tested through classification accuracy, sensitivity and specificity analysis, and confusion matrix analysis. The results obtained by AMMLP are compared with the backpropagation algorithm (BPA) and other recent classification techniques applied to the same database. The best result obtained so far with the AMMLP algorithm is 99.63%.

Software Sensor for Measuring Lactic Acid Concentration: Effect of Input Number and Node Number

Artificial Neural Network (ANN) approach was applied in developing software sensor for production of lactic acid using pineapple waste from Lactobacillus delbreuckii. Lactic acid production currently is one of the significant materials in industry and production with renewable source such as pineapple waste made the production of lactic acid faced a lot of disturbances in measuring the quality of lactic acid produced. An artificial neural network (ANN) was developed to predict the concentration of lactic acid, using collected data from an offline analysis. Multi layer perceptron (MLP) was used for mapping between the input and output parameters. Two variables were used as input parameters. MSE was used to evaluate the predictive performance of MLP. Logsig and purelin was used as the activation function and Levenberg-Marquadt was utilized as the training algorithm. The result showed that having 2 inputs is better in predicting the concentration of lactic acid; instead of 1 input. The optimum structure found was 2-5-1.

Estimation of surface shear strength in Zagros region of Iran — A comparison of artificial neural networks and multiple-linear regression models

Modeling soil erosion and runoff requires accurate estimates of surface soil shear strength. In this study, the use of artificial neural networks (ANNs) in estimating soil shear strength from measured particle size distribution, topographic attributes, normalized difference vegetation index (NDVI), soil organic carbon (SOC), and CaCO 3 were investigated. Three neural networks structures: multilayer perceptron (MLP), generalized feed-forward, and modular feed-forward networks were used and compared with conventional multiple-linear regression analysis. Improvements were achieved with all three neural network models over a conventional multiple-linear regression (MLR) models. Estimation model 1 of soil shear strength included particles size distribution, CaCO 3 and SOC; in model 2 particles size distribution, topographic attributes and NDVI were used; and in model 3 combination of model 1 and 2 as independent variables were used. The performances of the models were evaluated using Spearman's correlation coefficient ( r) between the observed and the estimated values, normalized mean square error (NMSE), and mean absolute error (MAE). The performance of MLP as compared to other ANN structures was significantly better ( p > 0.05). Model 3 (ANNs and regression) performed insignificantly better than model 2 (ANNs and regression) but model 2 and model 3 (ANNs and regression) were significantly different from model 1 (ANNs and regression). Slope and vegetation attributes were found to be the most sensitive variables to estimate surface soil shear strength in Zagros region of Iran.

Estimating impervious surfaces from medium spatial resolution imagery using the self-organizing map and multi-layer perceptron neural networks

The studies of impervious surfaces are important because they are related to many environmental problems, such as water quality, stream health, and the urban heat island effect. Previous studies have discussed that the self-organizing map (SOM) can provide a promising alternative to the multi-layer perceptron (MLP) neural networks for image classification at both per-pixel and sub-pixel level. However, the performances of SOM and MLP have not been compared in the estimation and mapping of urban impervious surfaces. In mid-latitude areas, plant phenology has a significant influence on remote sensing of the environment. When the neural networks approaches are applied, how satellite images acquired in different seasons impact impervious surface estimation of various urban surfaces (such as commercial, residential, and suburban/rural areas) remains to be answered. In this paper, an SOM and an MLP neural network were applied to three ASTER images acquired on April 5, 2004, June 16, 2001, and October 3, 2000, respectively, which covered Marion County, Indiana, United States. Six impervious surface maps were yielded, and an accuracy assessment was performed. The root mean square error (RMSE), the mean average error (MAE), and the coefficient of determination ( R 2) were calculated to indicate the accuracy of impervious surface maps. The results indicated that the SOM can generate a slightly better estimation of impervious surfaces than the MLP. Moreover, the results from three test areas showed that, in the residential areas, more accurate results were yielded by the SOM, which indicates that the SOM was more effective in coping with the mixed pixels than the MLP, because the residential area prevailed with mixed pixels. Results obtained from the commercial area possessed very high RMSE values due to the prevalence of shade, which indicates that both algorithms cannot handle the shade problem well. The lowest RMSE value was obtained from the rural area due to containing of less mixed pixels and shade. This research supports previous observations that the SOM can provide a promising alternative to the MLP neural network. This study also found that the impact of different map sizes on the impervious surface estimation is significant.

Regional TEC mapping using neural networks

Characterization and modeling of ionospheric variability in space and time is very important for communications and navigation. To characterize the variations, the ionosphere should be monitored, and the sparsity of the measurements has to be compensated by interpolation algorithms. The total electron content (TEC) is a major parameter that can be used to obtain regional ionospheric maps. In this study, neural networks (NNs), specifically multilayer perceptrons (MLPs) and radial basis function networks (RBFN), are investigated for the merits of their nonlinear modeling capability. In order to assess the performance of MLP and RBFN structures with respect to mapping and ionospheric parameters, these algorithms are applied to synthetically generated TEC surfaces representing various ionospheric states. Synthetic TEC data are sampled homogenously and randomly for a varying number of data points. The reconstruction errors show that the performance improves significantly when homogenous sampling is preferred to random station distribution. The best MLP and RBFN structures for any possible realistic scenario are determined by examining the performance parameters for all possible cases. It is also observed that RBFN with local receptive fields relies more on the number of training data points. In contrast to RBFN, MLP as a global approximator depends strongly on ionospheric trends. Finally, chosen MLP and RBFN models are applied to a set of real GPS‐TEC values obtained from central Europe, and their performances are compared with well known Global Ionospheric Maps produced by the International GNSS Service. The resolution and interpolation quality of the generated maps indicate that NNs offer a powerful and reliable alternative to the conventional TEC mapping algorithms.

The application of ridge polynomial neural network to multi-step ahead financial time series prediction

Motivated by the slow learning properties of multilayer perceptrons (MLPs) which utilize computationally intensive training algorithms, such as the backpropagation learning algorithm, and can get trapped in local minima, this work deals with ridge polynomial neural networks (RPNN), which maintain fast learning properties and powerful mapping capabilities of single layer high order neural networks. The RPNN is constructed from a number of increasing orders of Pi–Sigma units, which are used to capture the underlying patterns in financial time series signals and to predict future trends in the financial market. In particular, this paper systematically investigates a method of pre-processing the financial signals in order to reduce the influence of their trends. The performance of the networks is benchmarked against the performance of MLPs, functional link neural networks (FLNN), and Pi–Sigma neural networks (PSNN). Simulation results clearly demonstrate that RPNNs generate higher profit returns with fast convergence on various noisy financial signals.

Multilayer perceptron as inverse model in a ground-based remote sensing temperature retrieval problem

In this paper, a combustion temperature retrieval approximation for high-resolution infrared ground-based measurements has been developed based on a multilayer perceptron (MLP) technique. The introduction of a selection subset of features is mandatory due to the problems related to the high dimensionality data and the worse performance of MLPs with this high input dimensionality. Principal component analysis is used to reduce the input data dimensionality, selecting the physically important features in order to improve MLP performance. The use of a priori physical information over other methods in the chosen feature's phase has been tested and has appeared jointly with the MLP technique as a good alternative for this problem.

Use of Neural Networks for Automatic Classification From High-Resolution Images

The effectiveness of multilayer perceptron (MLP) networks as a tool for the classification of remotely sensed images has been already proven in past years. However, most of the studies consider images characterized by high spatial resolution (around 15-30 m) while a detailed analysis of the performance of this type of classifier on very high resolution images (around 1-2 m) such as those provided by the Quickbird satellite is still lacking. Moreover, the classification problem is normally understood as the classification of a single image while the capabilities of a single network of performing automatic classification and feature extraction over a collection of archived images has not been explored so far. In this paper, besides assessing the performance of MLP for the classification of very high resolution images, we investigate on the generalization capabilities of this type of algorithms with the purpose of using them as a tool for fully automatic classification of collections of satellite images, either at very high or at high-resolution. In particular, applications to urban area monitoring have been addressed

Optimized multilayer perceptrons for molecular classification and diagnosis using genomic data

Multilayer perceptrons (MLP) represent one of the widely used and effective machine learning methods currently applied to diagnostic classification based on high-dimensional genomic data. Since the dimensionalities of the existing genomic data often exceed the available sample sizes by orders of magnitude, the MLP performance may degrade owing to the curse of dimensionality and over-fitting, and may not provide acceptable prediction accuracy. Based on Fisher linear discriminant analysis, we designed and implemented an MLP optimization scheme for a two-layer MLP that effectively optimizes the initialization of MLP parameters and MLP architecture. The optimized MLP consistently demonstrated its ability in easing the curse of dimensionality in large microarray datasets. In comparison with a conventional MLP using random initialization, we obtained significant improvements in major performance measures including Bayes classification accuracy, convergence properties and area under the receiver operating characteristic curve (A(z)). The Supplementary information is available on http://www.cbil.ece.vt.edu/publications.htm

Tagging b jets associated with heavy neutral MSSM Higgs bosons

Since a neural network (NN) approach has been shown to be applicable to the problem of Higgs boson detection at LHC [I. Iashvili, A. Kharchilava, CMS TN-1996/100; M. Mjahed, Nucl. Phys. B 140 (2005) 799], we study the use of NNs in the problem of tagging b jets in pp → b b ¯ H SUSY , H SUSY → τ τ in the Compact Muons Solenoid experiment [F. Hakl, et al., Nucl. Instr. and Meth. A 502 (2003) 489; S. Lehti, CMS NOTE-2001/019; G. Segneri, F. Palla, CMS NOTE-2002/046]. B tagging is an important tool for separating the Higgs events with associated b jets from the Drell–Yan background Z , γ * → τ τ , for which the associated jets are mostly light quark and gluon jets. We teach multi-layer perceptrons (MLPs) available in the object oriented implementation of data analysis framework ROOT [ROOT—An Object Oriented Data Analysis Framework, in: Proceedings of the AIHENP’96 Workshop, Lausanne, September 1996, Nucl. Instr. and Meth. A 389 (1997) 81]. The following learning methods are evaluated: steepest descent algorithm, (BFGS) Broyden–Fletcher–Goldfarb–Shanno algorithm, and variants of conjugate gradients. The ROOT code generation feature of standalone C++ classifiers is utilized. We compare the b tagging performance of MLPs with another ROOT based feed forward NN tool NeuNet [J.P. Ernenwein, NeuNet software for ROOT], which uses a common back-propagation learning method. In addition, we demonstrate the use of the self-organizing map program package (SOM_PAK) and the learning vector quantization program package (LVQ_PAK) [T. Kohonen, et al., SOM_PAK: the self-organizing map program package, Technical Report A31; T. Kohonen, et al., LVQ_PAK: the learning vector quantization program package, Technical Report A30, Laboratory of Computer and Information Science, Helsinki University of Technology, FIN-02150 Espoo, Finland, 1996] in the b tagging problem.

Rainfall‐runoff modelling using artificial neural networks: comparison of network types

AbstractGrowing interest in the use of artificial neural networks (ANNs) in rainfall‐runoff modelling has suggested certain issues that are still not addressed properly. One such concern is the use of network type, as theoretical studies on a multi‐layer perceptron (MLP) with a sigmoid transfer function enlightens certain limitations for its use. Alternatively, there is a strong belief in the general ANN user community that a radial basis function (RBF) network performs better than an MLP, as the former bases its nonlinearities on the training data set. This argument is not yet substantiated by applications in hydrology. This paper presents a comprehensive evaluation of the performance of MLP‐ and RBF‐type neural network models developed for rainfall‐runoff modelling of two Indian river basins. The performance of both the MLP and RBF network models were comprehensively evaluated in terms of their generalization properties, predicted hydrograph characteristics, and predictive uncertainty. The results of the study indicate that the choice of the network type certainly has an impact on the model prediction accuracy. The study suggests that both the networks have merits and limitations. For instance, the MLP requires a long trial‐and‐error procedure to fix the optimal number of hidden nodes, whereas for an RBF the structure of the network can be fixed using an appropriate training algorithm. However, a judgment on which is superior is not clearly possible from this study. Copyright © 2004 John Wiley & Sons, Ltd.

Comparison of non-linear mixture models: sub-pixel classification

Sub-pixel level classification is essential for the successful description of many land cover patterns with spatial resolution of less than ~1 km and has been widely used in global or continental scale land cover mapping with remote sensing data. This paper presents a general comparison of four non-linear models for sub-pixel classification: ARTMAP, ART-MMAP, Regression Tree (RT) and Multilayer Perceptron (MLP) with Back-Propagation (BP) algorithm. The comparison is based on four factors: accuracy, model complexity, interpolation ability and error distribution. Two data sets, one simulated and one real world MODIS satellite image, were used to demonstrate the characteristics of each model. Experimental results show the superior performance of MLP with the simulated data set and better performance of ART-MMAP with the MODIS data set.

Temporal processing in primate motor control: relation between cortical and EMG activity.

We investigated spatio-temporal information processing in the primate motor system. Corticomotoneuronal (CM) cells provide monosynaptic excitatory connections from motor cortex to spinal motoneurons and contribute causally to the time-varying electromyogram (EMG) of their target muscle. A multilayer perceptron (MLP) was used to evaluate the transfer function between neural activity of single CM cells and their target muscle EMG, using data from in-vivo recordings in primate motor cortex. For an optimal MLP performance, i.e., minimal error between recorded target EMG and MLP-derived EMG, the CM cell input period had to span the latency observed between CM cell peak activity and EMG peak activity. We argue that the same spike train may code two types of information: 1) rate coding within the input window accounted for large-amplitude variations in the EMG signal and 2) temporal coding within a window of 40 ms just prior to the EMG output signal accounted for EMG variations of small amplitude. The transfer function of the MLP, thus, combines rate and temporal coding and suggests that CM cell output may also combine these two forms of coding. We predict that mutual constraints of rate and temporal coding would, however, would limit the CM output to code for particular temporal profiles of EMG, possibly adapted to bio-mechanical constraints.