Simple Multilayer Perceptron Research Articles

A novel neural Smith chart for use in microwave circuitry

In this article, briefly the Smith chart is mapped with an artificial neural network (ANN) covering its whole details to be exploited CAD of the microwave circuitry. Thus, relative to the similar works in the existing literature, this article provides the continuous Smith chart domain to facilitate the “Smith chart” methodology in solving the highly nonlinear transformation equations between the rectangular impedance and polar reflection planes for an infinite number of passive impedance to be used in design tasks of the microwave circuits. Data ensembles for the training and testing processes are obtained from the systematically selected locations on the Smith chart with the adaptive radius sampling algorithm. The ANN architecture is also simple, which consists of the two simple multilayer perceptron (MLP) modules with the common inputs which are the termination ZS = RS + jXS, line {ℓ, Z0} operation bandwidth B between the defined fmin, fmax and the dielectric ε. Briefly, the outputs of these ANN modules are the standing waves and the impedance transformation, which are the characteristic features of the transmission line circuits. Activation of the hidden layers of the modules are performed by the tangential-sigmoid type of function while the output layers are activated linearly. Furthermore, the neural unit element (NUE) is defined by the two independent neural networks as problems in the forward and reverse directions to be incorporated into the analysis and design algorithms of the unit element (UE). This can also be considered as solving the simultaneous nonlinear equation set for (ℓ, Z0) parameters of the required impedance transformations ZOUT(ω) = ROUT (ω) + XOUT (ω) from the given complex termination ZS = RS + jXS. Applications of the Neural Smith chart are given by the numerous examples with the proved accuracy. Thus it has been verified that this neural Smith chart can be exploited for the whole classical transmission line theory including impedance matching. © 2008 Wiley Periodicals, Inc. Int J RF and Microwave CAE, 2009.

Foreground removal from CMB temperature maps using an MLP neural network

One of the main obstacles for extracting the Cosmic Microwave Background (CMB) signal from observations in the mm-submm range is the foreground contamination by emission from Galactic components: mainly synchrotron, free-free and thermal dust emission. Due to the statistical nature of the intrinsic CMB signal it is essential to minimize the systematic errors in the CMB temperature determinations.Following the available knowledge of the spectral behavior of the Galactic foregrounds simple power law-like spectra have been assumed. The feasibility of using a simple neural network for extracting the CMB temperature signal from the combined signal CMB and the foregrounds has been investigated. As a specific example, we have analysed simulated data, as expected from the ESA Planck CMB mission. A simple multilayer perceptron neural network with 2 hidden layers can provide temperature estimates over more than 80 per cent of the sky that are to a high degree uncorrelated with the foreground signals. A single network will be able to cover the dynamic range of the Planck noise level over the entire sky.

Fast and accurate modelling of frequency-selective surfaces using a new modular neural network configuration of multilayer perceptrons

The authors consider the analysis and modelling of the scattering from frequency-selective surfaces (FSSs), in the 6–14 GHz band, as a function of its periodic array geometry of thin dipole elements on an anisotropic layer. The accurate full-wave electromagnetic (EM) analysis of each FSS was carried out using the method of moments. From the available EM data, the artificial neural network (ANN) models can be developed. The modelling problem was solved by using a new modular configuration of multilayer perceptrons (MLPs), which is an implementation of the proposal modified from the previous knowledge method of neuromodelling information. Each MLP in the modular configuration was trained separately from the others through the resilient backpropagation algorithm. Within the region of interest studied, the ANN model developed is able to estimate the resonance frequencies and the bandwidths of the FSS band-stop filters, with high accuracy and low computational cost. To verify the advantageous properties of the modular MLP/MLP model, a neural model using a simple MLP was developed in order to analyse the same learning task. A comparative study was done between these models in terms of training the convergence, the accuracy and the computational cost.

A Mixture of Multilayer Perceptron Experts Network for Modeling Face/Nonface Recognition in Cortical Face Processing Regions

Abstract Recent studies in neurobiology and especially in neuroimaging report that a gating mechanism prior to face processing levels of human visual system, facilitates the face/nonface recognition task. In accordance to these biological evidences, we propose aface/nonface recognition model which makes use of mixture of experts network. In order to improve the face/nonface recognition accuracy, the outputs of the expert networks aze combined using a gating network. A novel structure, which is the use of multilayer perceptrons (MLPs) in forming the expert networks, is introduced. The learning algorithm is modified to be adapted with the MLP networks. The results reveal that using a mixture of simple MLPs is much more beneficial, in many respects, as it shows more certainty at its output and is also easier to train than a single, but complex, MLP.

An assessment of machine and statistical learning approaches to inferring networks of protein-protein interactions

Abstract Protein-protein interactions (PPI) play a key role in many biological systems. Over the past few years, an explosion in availability of functional biological data obtained from high-throughput technologies to infer PPI has been observed. However, results obtained from such experiments show high rates of false positives and false negatives predictions as well as systematic predictive bias. Recent research has revealed that several machine and statistical learning methods applied to integrate relatively weak, diverse sources of large-scale functional data may provide improved predictive accuracy and coverage of PPI. In this paper we describe the effects of applying different computational, integrative methods to predict PPI in Saccharomyces cerevisiae. We investigated the predictive ability of combining different sets of relatively strong and weak predictive datasets. We analysed several genomic datasets ranging from mRNA co-expression to marginal essentiality. Moreover, we expanded an existing multi-source dataset from S. cerevisiae by constructing a new set of putative interactions extracted from Gene Ontology (GO)- driven annotations in the Saccharomyces Genome Database. Different classification techniques: Simple Naive Bayesian (SNB), Multilayer Perceptron (MLP) and K-Nearest Neighbors (KNN) were evaluated. Relatively simple classification methods (i.e. less computing intensive and mathematically complex), such as SNB, have been proven to be proficient at predicting PPI. SNB produced the “highest” predictive quality obtaining an area under Receiver Operating Characteristic (ROC) curve (AUC) value of 0.99. The lowest AUC value of 0.90 was obtained by the KNN classifier. This assessment also demonstrates the strong predictive power of GO-driven models, which offered predictive performance above 0.90 using the different machine learning and statistical techniques. As the predictive power of single-source datasets became weaker MLP and SNB performed better than KNN. Moreover, predictive performance saturation may be reached independently of the classification models applied, which may be explained by predictive bias and incompleteness of existing “Gold Standards”. More comprehensive and accurate PPI maps will be produced for S. cerevisiae and beyond with the emergence of largescale datasets of better predictive quality and the integration of intelligent classification methods.

Excessive Noise Injection Training of Neural Networks for Markerless Tracking in Obscured and Segmented Environments

In this letter, we demonstrate that the generalization properties of a neural network (NN) can be extended to encompass objects that obscure or segment the original image in its foreground or background. We achieve this by piloting an extension of the noise injection training technique, which we term excessive noise injection (ENI), on a simple feedforward multilayer perceptron (MLP) network with vanilla backward error propagation to achieve this aim. Six tests are reported that show the ability of an NN to distinguish six similar states of motion of a simplified human figure that has become obscured by moving vertical and horizontal bars and random blocks for different levels of obscuration. Four more extensive tests are then reported to determine the bounds of the technique. The results from the ENI network were compared to results from the same NN trained on clean states only. The results pilot strong evidence that it is possible to track a human subject behind objects using this technique, and thus this technique lends itself to a real-time markerless tracking system from a single video stream.

Application of neural networks for extraction of distance and reflectance in pulsed laser radar

In this paper a new method has been developed whereby distance and reflectance information can be obtained directly from the output waveform of pulsed laser radar (PLR). A simple digital signal processing technique and multilayer perceptrons (MLP) type neural networks have been used to recognize the pulse shapes and to obtain the information. The method has been applied to PLR for three purposes: high-resolution distance measurement, very short distance measurement and two-dimensional (2D) imaging. In order to present a precise evaluation, all measurements were performed simultaneously by the new approach and the standard method of pulsed time of flight, which is very often used for extraction of information from the output signal of PLR system. In comparison with the standard method, in the high-resolution distance measurement applications, the non-linearity errors are lowered by more than 6% and the standard deviation of the single-shot distance measurement results is lowered by almost one order. In very short distance measurement applications, for the first time distance measurements with overlapping of the reference and reflected pulses have been possible with no loss in distance resolution. In the reflectance measurement applications (2D imaging), the average error and maximum error in the reflectance measurements have been improved by 88.5% and 72.6%, respectively. In all cases, in addition to these improvements, since the new method decreases the effect of the noise, it is possible to make the measurements with the same resolution as the standard method, but with a lower averaging in the sampling unit thus drastically increasing the speed of the measurements.

Part 2: Multilayer perceptron and natural gradient learning

Since the perceptron was developed for learning to classify input patterns, there have been plenty of studies on simple perceptrons and multilayer perceptrons. Despite wide and active studies in theory and applications, multilayer perceptrons still have many unsettled problems such as slow learning speed and overfitting. To find a thorough solution to these problems, it is necessary to consolidate previous studies, and find new directions for uplifting the practical power of multilayer perceptrons. As a first step toward the new stage of studies on multilayer perceptrons, we give short reviews on two interesting and important approaches; one is stochastic approach and the other is geometric approach. We also explain an efficient learning algorithm developed from the statistical and geometrical studies, which is now well known as the natural gradient learning method.

Genetic engineering of hierarchical fuzzy regional representations for handwritten character recognition

This paper presents a genetic programming based approach for optimizing the feature extraction step of a handwritten character recognizer. This recognizer uses a simple multilayer perceptron as a classifier and operates on a hierarchical feature space of orientation, curvature, and center of mass primitives. The nodes of the hierarchy represent rectangular sub-regions of their parent node, the tree root corresponding to the character's bounding box. Within each sub-region, a variable number of fuzzy features are extracted. Genetic programming is used to simultaneously learn the best hierarchy and the best combination of fuzzy features. Moreover, the fuzzy features are not predetermined, they are inferred from the evolution process which runs a two-objective selection operator. The first objective maximizes the recognition rate, and the second minimizes the feature space size. Results on Unipen data show that, using this approach, robust representations could be obtained that out-performed comparable human designed hierarchical fuzzy regional representations.

Dualband FLIR fusion for automatic target recognition

We investigate the potential benefits of fusing two bands of forward-looking infrared (FLIR) data for target detection and clutter rejection. We propose a similar set of neural-based clutter rejecters and target detectors, each of which consists of an eigenspace transformation and a simple multilayer perceptron. The same architecture is used to operate on either single band or dualband FLIR input images, so that the net effects of dualband fusion can be demonstrated. When the dualband inputs are used, the component bands are combined at either pixel or feature level, thus providing insight into methods of performing data fusion in this particular application. A large set of real FLIR images is used in two series of experiments, one for clutter rejection tasks and the other for target detection tasks. In both series, the results indicate that the dualband input images do improve the performance of the clutter rejecters and target detectors over their single band counterparts. On the other hand, results of the pixel and feature level fusions are quite similar, suggesting that dimensionality reduction by the eigenspace transformation can be performed independently on the two bands.

Neural speed control for autonomous road vehicles

In this paper a neural-network automotive speed controller is presented which works at low and at high speed-levels using throttle and brake control input. It can be used for autonomous intelligent cruise control including stop-and-go traffic situations. The network itself consists of a simple multilayer feedforward perceptron network. A special training method is used, where the neural network is trained on a detailed nonlinear dynamic vehicle model. Practical road tests with the Daimler Benz experimental vehicle OSCAR (MB 300 TE station wagon) show good results for both high- and low-speed control.

Neural network implementation using a single MOST per synapse

A VLSI implementation of an artificial neural network using a single n-channel MOS (metal-oxide semiconductor) transistor per synapse is investigated. The simplicity of the design is achieved by using pulse width modulation to represent neural activity and by using a novel technique to manipulate negative weights. A simple multilayer perceptron (MLP) network was simulated using the SPICE circuit simulator and the performance of a hardware realization of the same MLP network was measured. Simulations and measurements are shown to agree well.

Neural Speed Control for Autonomous Road Vehicles

In this paper a neural network automotive speed controller is presented which works at low and at high speed-levels using throttle and brake control input. It can be used for autonomous intelligent cruise control including stop & go traffic situations. The network itself consists of a simple multilayer feedforward perceptron network. A special training method is used where the neural network is trained on a detailed nonlinear dynamical vehicle model. Practical roadtests with the Daimler Benz experimental vehicle OSCAR (MB 300 TE station wagon) show good results for both high and low speed control.

Neural networks and discrimination of seismic signals

Recent developments in algorithms and computer architecture make neural networks a useful tool in designing pattern recognition systems. We show how a simple multilayer perceptron with 23 neurons can be trained easily and used to classify seismic signals. Applied to broadband seismic signal, the perceptron permitted the recognition of different types of events on the basis of their frequency. Applied to a real-time, automatic, seismic data acquisition system, it saved more than 50% CPU time in a detection procedure.

A neural network classifier in experimental particle physics.

A classification problem in high energy physics has been solved on simulated data using a simple multilayer perceptron comprising binary units which was trained with the CHIR algorithm. The unstable training of such a network on a nonseparable set has been overcome by selecting those weight vectors with good performance while providing a flexible choice of the two types of classification errors. Specific features of the problem have been exploited in order to simplify and optimize the solution which has been compared to the popular backpropagation algorithm and found to perform on a similar level. Additional aspects of this work are the use of the CHIR algorithm on continuous input and incorporating the classic idea of a phi-machine in a multilayer perceptron.

University of Sussex

The use of artificial intelligence methods in different hydraulic sciences has become conventional in recent years. In this study, two artificial neural networks (ANN), namely multilayer perceptron (MLP) and radial basis function (RBF) models were modified with decision trees (DT) and designed as two new hybrid models, namely DT-MLP and DT-RBF. The performance of the proposed hybrid (DT-MLP and DT-RBF) and simple (MLP and RBF) models was compared for velocity and water surface prediction in sharp 90° bends. The experimental data for 5 different hydraulic conditions in a sharp 90° bend were used to train and test the models. In velocity prediction, the mean absolute error (MAE), root mean square error (RMSE) and relative error (δ) values decreased with the DT-MLP model compared to the MLP model by 16%, 9% and 0.17% and the values of DT-RBF reduced by 11%, 7% and 0.11% compared to the RBF model, respectively. For water surface prediction, the MAE, RMSE and δ errors decreased with the DT-MLP model by 5.2%, 5.5% and 0.095% compared with MLP and with the DT-RBF model the errors decreased by 20%, 23% and 0.5% compared with RBF, respectively. Using the new proposed hybrid algorithms based on decision trees enhanced the simple MLP and RBF models’ performance.

Relation between brinell ball tests and scleroscope readings : J. J. Thomas. ( Amer. Soc. Testing Materials, 1914)

In this paper a neural-network automotive speed controller is presented which works at low and at high speed-levels using throttle and brake control input. It can be used for autonomous intelligent cruise control including stop-and-go traffic situations. The network itself consists of a simple multilayer feedforward perceptron network. A special training method is used, where the neural network is trained on a detailed nonlinear dynamic vehicle model. Practical road tests with the Daimler Benz experimental vehicle OSCAR (MB 300 TE station wagon) show good results for both high- and low-speed control.