Search

Prediction of machine tool failure has been very important in modern metal cutting operations in order to meet the growing demand for product quality and cost reduction. This paper presents the study of building a neural network model for predicting the behavior of a boring process during its full life cycle. This prediction is achieved by the fusion of the predictions of three principal components extracted as features from the joint time–frequency distributions of energy of the spindle loads observed during the boring process. Furthermore, prediction uncertainty is assessed using nonlinear regression in order to quantify the errors associated with the prediction. The results show that the implemented Elman recurrent neural network is a viable method for the prediction of the feature behavior of the boring process, and that the constructed confidence bounds provide information crucial for subsequent maintenance decision making based on the predicted cutting tool degradation. [ABSTRACT FROM AUTHOR]

The concept design of checking fixtures for auto-body parts is a highly complex process that requires a human designer to draw from his rich experience. By exploiting recent advances in CAD/CAM and artificial intelligence techniques, one may constrain multiple solutions such that only good designs are considered. In this paper, a method of selecting type for checking fixtures is proposed that harnesses advantages of neural networks. This method attempts to capture relevant domain knowledge and is used to produce acceptable solutions. The method is applied to a case problem and the suggested checking fixture type is compared to one offered by a human designer. The agreement between the two solutions is very close. [ABSTRACT FROM AUTHOR]

In this paper, an “intelligent” isolated intersection control system was developed. The developed “intelligent” system makes “real time” decisions as to whether to extend (and how much) current green time. The model developed is based on the combination of the dynamic programming and neural networks. Many tests show that the outcome (the extension of the green time) of the proposed neural network is nearly equal to the best solution. Practically negligible CPU times were achieved, and were thus absolutely acceptable for the “real time” application of the developed algorithm. [ABSTRACT FROM AUTHOR]

Condition monitoring of machine tool inserts is important for increasing the reliability and quality of machining operations. Various methods have been proposed for effective tool condition monitoring (TCM), and currently it is generally accepted that the indirect sensor-based approach is the best practical solution to reliable TCM. Furthermore, in recent years, neural networks (NNs) have been shown to model successfully, the complex relationships between input feature sets of sensor signals and tool wear data. NNs have several properties that make them ideal for effectively handling noisy and even incomplete data sets. There are several NN paradigms which can be combined to model static and dynamic systems. Another powerful method of modeling noisy dynamic systems is by using hidden Markov models (HMMs), which are commonly employed in modern speech-recognition systems. The use of HMMs for TCM was recently proposed in the literature. Though the results of these studies were quite promising, no comparative results of competing methods such as NNs are currently available. This paper is aimed at presenting a comparative evaluation of the performance of NNs and HMMs for a TCM application. The methods are employed on exactly the same data sets obtained from an industrial turning operation. The advantages and disadvantages of both methods are described, which will assist the condition-monitoring community to choose a modeling method for other applications. [ABSTRACT FROM AUTHOR]

Some medical and epidemiological surveys have been designed to predict a nominal response variable with several levels. With regard to the type of pregnancy there are four possible states: wanted, unwanted by wife, unwanted by husband and unwanted by couple. In this paper, we have predicted the type of pregnancy, as well as the factors influencing it using two different models and comparing them. Regarding the type of pregnancy with several levels, we developed a multinomial logistic regression and a neural network based on the data and compared their results using three statistical indices: sensitivity, specificity and kappa coefficient. Based on these three indices, neural network proved to be a better fit for prediction on data in comparison to multinomial logistic regression. When the relations among variables are complex, one can use neural networks instead of multinomial logistic regression to predict the nominal response variables with several levels in order to gain more accurate predictions. [ABSTRACT FROM AUTHOR]

The grade of textile yarns is an important index in evaluating the yarn’s market value. This paper uses the backpropagation neural network (BNN) and Karhunen-Loeve (K-L) expansion method to construct a new and highly accurate grading system. Outcomes show that a highly accurate and neutral grading system can be obtained if the BNN learning sample is comprehensive or by adopting the BNN with a relearning technique (self-healing). Considering the possibility of reducing the dimension of BNN input vectors without losing the accuracy, this paper preprocesses the BNN grading system using the K-L expansion. Experiments demonstrate that the K-L expansion provides a way to reduce the input dimensions, and that a single principle axis value of the BNN with the K-L expansion grading system is able to grade textile yarns. In addition, the experiment demonstrates that as the input dimensions are reduced to four in a self-healing neural network with the K-L expansion, the grading system provides the high accuracy and robustness. [ABSTRACT FROM AUTHOR]

A number of firm performance models are available. Reviewing these models and pointing out their individual strengths and weaknesses, would help both academic researchers and professional users to understand and appreciate how and when to use these various models. The theoretical models for Small and Medium-size Enterprise (SME) performance can be divided into two categories: firm dynamics theories and performance prediction models. In the first part of this paper we review, in a condensed manner, the most relevant firm dynamic theories, i.e. SME's performance models. These include: Stochastic Theories, Learning Model Theories and Hazard Modeling Theories. In the second part of this paper, we examine the performance prediction models of SMEs, which include Z-Scores, ZETA-Scores, Neural Networks (NN) and the SIV® models, among others. The strengths and weaknesses of each of these models are exposed and discussed. [ABSTRACT FROM AUTHOR]

Solar energetic particle events can cause major disruptions to the operation of spacecraft in earth orbit and outside the earth's magnetosphere and have to be considered for EVA and other manned activities. They may also have an effect on radiation doses received by the crew flying in high altitude aircraft over the polar regions. The occurrence of these events has been assumed to be random, but there would appear to be some solar cycle dependency with a higher annual fluence occuring during a 7 year period, 2 years before and 4 years after the year of solar maximum. Little has been done to try to predict these events in real-time with nearly all of the work concentrating on statistical modelling. Currently our understanding of the causes of these events is not good. But what are the prospects for prediction? Can artificial intelligence techniques be used to predict them in the absence of a more complete understanding of the physics involved? The paper examines the phenomenology of the events, briefly reviews the results of neural network prediction techniques and discusses the conjecture that the underlying physical processes might be related to self-organised criticality and turblent MHD flows. [ABSTRACT FROM AUTHOR]

The purpose of the paper is the evaluation of a radial basis function neural network as a tool for computer aided coronary artery disease diagnosis based on the results of the traditional ECG exercise test. The research was performed using 776 data records from an exercise test (297 records from healthy patients and 479 from ill patients) confirmed by coronary arteriography results. Each record described the state of the patient, provided input data for the neural network, included the level and slope of an ST segment of a 12-lead ECG signal made at rest and after effort, heart rate, blood pressure, load during the test, and occurrence of coronary pain, coronary arteriography, correct output pattern for the neural network, and verified the existence (or not) of more than 50% stenosis of the particular coronary vessels. Radial basis function neural networks for coronary artery disease diagnosis were optimised by choosing the type of radial function, the method of training (setting the number of centres and their dimensions), and regularisation. The best network correctly recognised over 97% of cases from a 400-element test set, diagnosing not only the patients' condition (simple 'healthy/unhealthy' diagnosis), but also pointing out individual unhealthy/stenosed vessels. [ABSTRACT FROM AUTHOR]

Conventional supervised learning in neural networks is carried out by performing unconstrained minimization of a suitably defined cost function. This approach has certain drawbacks, which can be overcome by incorporating additional knowledge in the training formalism. In this paper, two types of such additional knowledge are examined: Network specific knowledge (associated with the neural network irrespectively of the problem whose solution is sought) or problem specific knowledge (which helps to solve a specific learning task). A constrained optimization framework is introduced for incorporating these types of knowledge into the learning formalism. We present three examples of improvement in the learning behaviour of neural networks using additional knowledge in the context of our constrained optimization framework. The two network specific examples are designed to improve convergence and learning speed in the broad class of feedforward networks, while the third problem specific example is related to the efficient factorization of 2-D polynomials using suitably constructed sigma-pi networks. [ABSTRACT FROM AUTHOR]

The forecasting and modeling of ionospheric scintillation effects are crucial for precise satellite positioning and navigation applications. In this paper, a Neural Network model, trained using Particle Swarm Optimization (PSO) algorithm, has been implemented for the prediction of amplitude scintillation index (S4) observations. The Global Positioning System (GPS) and Ionosonde data available at Darwin, Australia (12.4634° S, 130.8456° E) during 2013 has been considered. The correlation analysis between GPS S4 and Ionosonde drift velocities (hmf2 and fof2) data has been conducted for forecasting the S4 values. The results indicate that forecasted S4 values closely follow the measured S4 values for both the quiet and disturbed conditions. The outcome of this work will be useful for understanding the ionospheric scintillation phenomena over low latitude regions. [ABSTRACT FROM AUTHOR]

Ordinal regression is a kind of regression analysis used for predicting an ordered response variable. In these problems, the patterns are labelled by a set of ranks with an ordering among the different categories. The most common type of ordinal regression model is the cumulative link model. The cumulative link model relates an unobserved continuous latent variable with a monotone link function. Logit and probit functions are examples of link functions used in cumulative link models. In this paper, a novel generalized link function based on a generalization of the logistic distribution is proposed. The generalized link function proposed is able to reproduce other different link functions by changing two real parameters: $$\alpha $$ and $$\lambda $$ . The generalized link function has been included in a cumulative link model where the latent function is determined by a standard neural network in order to test the performance of the proposal. For this model, a reformulation of the tunable thresholds and distribution parameters was applied to convert the constrained optimization problem into an unconstrained optimization problem. Experimental results demonstrate that our proposed approach can achieve competitive generalization performance. [ABSTRACT FROM AUTHOR]

In this paper, an interpolation neural network is introduced for the learning of a wind turbine behavior with incomplete data. The proposed hybrid method is the combination of an interpolation algorithm and a neural network. The interpolation algorithm is applied to estimate the missing data of all the variables; later, the neural network is employed to learn the output behavior. The proposed method avoids the requirement to know all the system data. Experiments show the effectiveness of the proposed technique. [ABSTRACT FROM AUTHOR]

Wireless sensor networks (WSNs) are spatially distributed devices to support various applications. The undesirable behavior of the sensor node affects the computational efficiency and quality of service. Fault detection, identification, and isolation in WSNs will increase assurance of quality, reliability, and safety. In this paper, a novel neural network based fault diagnosis algorithm is proposed for WSNs to handle the composite fault environment. Composite fault includes hard, soft, intermittent, and transient faults. The proposed fault diagnosis protocol is based on gradient descent and evolutionary approach. It detects, diagnose, and isolate the faulty nodes in the network. The proposed protocol works in four phases such as clustering phase, communication phase, fault detection and classification phase, and isolation phase. Simulation results show that the proposed protocol performs better than the existing protocols in terms of detection accuracy, false alarm rate, false positive rate, and detection latency. [ABSTRACT FROM AUTHOR]

Volatile organic compounds refer to a large class of carbon-based chemicals capable of evaporating easily into the air at room temperature. Formaldehyde is one of the best known volatile organic compounds, and long-term exposure to formaldehyde emission from wood-based building products in indoor air may cause many adverse health effects. This paper presents an implementation of artificial neural networks for modeling the formaldehyde emission from particleboard as a wood-based product based on wood-glue moisture content, density of board and pressing temperature, with the experimental data collected from Petinarakis and Kavvouras (Wood Res 51(1):31-40, 2006). With the constructed model, formaldehyde emission of particleboard could be predicted successfully, and the intermediate formaldehyde emission values not obtained from experimental investigation could be predicted for different combinations of manufacturing parameters. The results proved that the artificial neural network is a promising technique in predicting the formaldehyde emission from particleboard. In this regard, the findings of this study will help the manufacturing industries in obtaining the intermediate values of the formaldehyde emission without performing further experimental activity. The model thus may save time, reduce the consumption of experimental materials and design costs. [ABSTRACT FROM AUTHOR]

The impending end of Moore's Law has started a rethinking of the way computers are built and computation is done. This paper discusses two directions that are currently attracting much attention as future computation paradigms: the merging of logic and memory, and brain-inspired computing. Natural computing has been known for its innovative methods to conduct computation, and as such may play an important role in the shaping of the post-Moore era. [ABSTRACT FROM AUTHOR]

In our days, due the evolution of high-speed computers, the old Human-Computer Interface (HCI) legacies based on mouse and keyboard are slowly becoming obsolete and cannot be accurate enough and respond in a timely manner to the flow of information today. This is why new ways of communicating with the computer have to be researched, the most natural one being the use of gestures. In this paper, a two-level architecture for recognizing human gestures from video frames is proposed. The architecture makes use of several feed-forward neural networks to compute the gestures based on the Haar-like features of body, hand and finger as well as a stochastic-free context grammar that is employed to comprise the mutual context between body pose and hand movement. Trained and tested on 10 gestures (Swipe Right, Swipe Left, Swipe Up, Swipe Down, Horizontal Wave, Vertical Wave, Circle, Point, Palm Up and Fist) the over 94 % accuracy of the system surpasses the current state of the art and compared with a system with no mutual context between body position and hand movement our proposed architecture shows an increase in accuracy with up to 7 %. [ABSTRACT FROM AUTHOR]

Overlapped fingerprints are often found in latent fingerprints lifted from crime scenes and in live-scan fingerprint images when the surface of fingerprint sensors contains residues of fingerprints of previous users. Such overlapped fingerprints usually cannot be processed accurately by contemporary commercial fingerprint matchers, which has led many researchers to propose methods designed to separate the overlapped fingerprints. In this paper, we propose a novel latent overlapped fingerprints separation algorithm based on neural networks. Our algorithm works in a block-based fashion. After producing an initial estimation of the orientation fields present in the overlapped fingerprint image, it uses a neural network to separate the mixed orientation fields, which are then post-processed to correct remaining errors and enhanced using the global orientation field enhancement model. Experimental results show that the proposed algorithm outperforms the state-of-the-art algorithm in terms of accuracy on the Tsinghua Overlapped Latent Fingerprint Database (containing real-world overlapped fingerprints obtained by forensic methods), while also showing encouraging results (second only to state-of-the-art) on the Tsinghua Simulated Overlapped Fingerprint Database (containing artificially overlapped fingerprints of a good quality). [ABSTRACT FROM AUTHOR]

Background: Online consumer reviews have become a baseline for new consumers to try out a business or a new product. The reviews provide a quick look into the application and experience of the business/product and market it to new customers. However, some businesses or reviewers use these reviews to spread fake information about the business/product. The fake information can be used to promote a relatively average product/business or can be used to malign their competition. This activity is known as reviewer fraud or opinion spam. The paper proposes a feature set, capturing the user social interaction behavior to identify fraud. The problem being solved is one of the characteristics that lead to fraud rather than detecting fraud. Methods: Neural network algorithm is used to evaluate the proposed feature set and compare it against the state-of-the-art feature sets in detecting fraud. The feature set considers the user's social interaction on the Yelp platform to determine if the user is committing fraud. The neural network algorithm helps in comparing the feature set with other feature sets used to detect fraud. Any attempt to find the characteristics that lead to fraud has a prerequisite to be good enough to detect fraud as well. Results: The F1 score obtained using neural networks is on par with all the well-known methods for detecting fraud, a value of 0.95. The effectiveness of the feature set is in rivaling the other approaches to fraud detection. Conclusions: A user's social interaction on a digital platform such as Yelp is equally important in evaluating the user as social interaction is in real life. The characteristics that lead to fraud can be intuitively captured. The characteristics such as number of friends, number of followers and the number of times the user has provided a review which was helpful to multiple people provide the neural network with a base to form a relationship between opinion fraud and social interaction characteristics. [ABSTRACT FROM AUTHOR]

Today artificial neural networks are applied in various fields-engineering, data analysis, robotics. While they represent a successful tool for a variety of relevant applications, mathematically speaking they are still far from being conclusive. In particular, they suffer from being unable to find the best configuration possible during the training process (local minimum problem). In this paper, we focus on this issue and suggest a simple, but effective, post-learning strategy to allow the search for improved set of weights at a relatively small extra computational cost. Therefore, we introduce a novel technique based on analogy with quantum effects occurring in nature as a way to improve (and sometimes overcome) this problem. Several numerical experiments are presented to validate the approach. [ABSTRACT FROM AUTHOR]

Active Contour Models ( ACMs) constitute an efficient energy-based image segmentation framework. They usually deal with the segmentation problem as an optimization problem, formulated in terms of a suitable functional, constructed in such a way that its minimum is achieved in correspondence with a contour that is a close approximation of the actual object boundary. However, for existing ACMs, handling images that contain objects characterized by many different intensities still represents a challenge. In this paper, we propose a novel ACM that combines-in a global and unsupervised way-the advantages of the Self-Organizing Map ( SOM) within the level set framework of a state-of-the-art unsupervised global ACM, the Chan-Vese (C-V) model. We term our proposed model SOM-based Chan-Vese ( SOMCV) active contour model. It works by explicitly integrating the global information coming from the weights (prototypes) of the neurons in a trained SOM to help choosing whether to shrink or expand the current contour during the optimization process, which is performed in an iterative way. The proposed model can handle images that contain objects characterized by complex intensity distributions, and is at the same time robust to the additive noise. Experimental results show the high accuracy of the segmentation results obtained by the SOMCV model on several synthetic and real images, when compared to the Chan-Vese model and other image segmentation models. [ABSTRACT FROM AUTHOR]

In this paper, the exponential stability of anti-periodic solutions for inertial neural networks with time delays is investigated. First, by properly chosen variable substitution the system is transformed to first order differential equation. Second, some sufficient conditions which can ensure the existence and exponential stability of anti-periodic solutions for the system are obtained by using Lyapunov method and uniformly converges. Finally, an example is given to illustrate the effectiveness of the results. [ABSTRACT FROM AUTHOR]

Soccer match attendance is an example of group behavior with noisy context that can only be approximated by a limited set of quantifiable factors. However, match attendance is representative of a wider spectrum of context-based behaviors for which only the aggregate effect of otherwise individual decisions is observable. Modeling of such behaviors is desirable from the perspective of economics, psychology, and other social studies with prospective use in simulators, games, product planning, and advertising. In this paper, we evaluate the efficiency of different neural network architectures as models of context in attendance behavior by comparing the achieved prediction accuracy of a multilayer perceptron (MLP), an Elman recurrent neural network (RNN), a time-lagged feedforward neural network (TLFN), and a radial basis function network (RBFN) against a multiple linear regression model, an autoregressive moving average model with exogenous inputs, and a naive cumulative mean model. We show that the MLP, TLFN, and RNN are superior to the RBFN and achieve comparable prediction accuracy on datasets of three teams from the English Football League Championship, which indicates weak importance of context transition modeled by the TLFN and the RNN. The experiments demonstrate that all neural network models outperform linear predictors by a significant margin. We show that neural models built on individual datasets achieve better performance than a generalized neural model constructed from pooled data. We analyze the input parameter influences extracted from trained networks and show that there is an agreement between nonlinear and linear measures about the most significant attributes. [ABSTRACT FROM AUTHOR]

The extension of estimation of distribution algorithms (EDAs) to the multi-objective domain has led to multi-objective optimization EDAs (MOEDAs). Most MOEDAs have limited themselves to porting single-objective EDAs to the multi-objective domain. Although MOEDAs have proved to be a valid approach, the last point is an obstacle to the achievement of a significant improvement regarding 'standard' multi-objective optimization evolutionary algorithms. Adapting the model-building algorithm is one way to achieve a substantial advance. Most model-building schemes used so far by EDAs employ off-the-shelf machine learning methods. However, the model-building problem has particular requirements that those methods do not meet and even evade. The focus of this paper is on the model-building issue and how it has not been properly understood and addressed by most MOEDAs. We delve down into the roots of this matter and hypothesize about its causes. To gain a deeper understanding of the subject we propose a novel algorithm intended to overcome the drawbacks of current MOEDAs. This new algorithm is the multi-objective neural estimation of distribution algorithm (MONEDA). MONEDA uses a modified growing neural gas network for model-building (MB-GNG). MB-GNG is a custom-made clustering algorithm that meets the above demands. Thanks to its custom-made model-building algorithm, the preservation of elite individuals and its individual replacement scheme, MONEDA is capable of scalably solving continuous multi-objective optimization problems. It performs better than similar algorithms in terms of a set of quality indicators and computational resource requirements. [ABSTRACT FROM AUTHOR]

The purpose of this paper is to critically compare a neural network technique with the established statistical technique of logistic regression for modeling decisions for several marketing situations. In our study, these two modeling techniques were compared using data collected on the decision by supermarket buyers whether to add a new product to their shelves or not. Our analysis shows that although neural networks offer a possible alternative approach, they have both strengths and weaknesses that must be clearly understood. [ABSTRACT FROM AUTHOR]

Air pollution stands as a significant modern-day challenge impacting life quality, the environment, and the economy. It comprises various pollutants like gases, particulate matter, biological molecules, and more, stemming from sources such as vehicle emissions, industrial operations, agriculture, and natural events. Nitrogen dioxide (NO2), among these harmful gases, is notably prevalent in densely populated urban regions. Given its adverse effects on health and the environment, accurate monitoring of NO2 levels becomes imperative for devising effective risk mitigation strategies. However, the precise measurement of NO2 poses challenges as it traditionally relies on costly and bulky equipment. This has prompted the development of more affordable alternatives, although their reliability is often questionable. The aim of this article is to introduce a groundbreaking method for precisely calibrating cost-effective NO2 sensors. This technique involves statistical preprocessing of low-cost sensor readings, aligning their distribution with reference data. Central to this calibration is an artificial neural network (ANN) surrogate designed to predict sensor correction coefficients. It utilizes environmental variables (temperature, humidity, atmospheric pressure), cross-references auxiliary NO2 sensors, and incorporates short time series of previous readings from the primary sensor. These methods are complemented by global data scaling. Demonstrated using a custom-designed cost-effective monitoring platform and high-precision public reference station data collected over 5 months, every component of our calibration framework proves crucial, contributing to its exceptional accuracy (with a correlation coefficient near 0.95 concerning the reference data and an RMSE below 2.4 µg/m3). This level of performance positions the calibrated sensor as a viable, cost-effective alternative to traditional monitoring approaches. [ABSTRACT FROM AUTHOR]

Neural network force fields have significantly advanced ab initio atomistic simulations across diverse fields. However, their application in the realm of magnetic materials is still in its early stage due to challenges posed by the subtle magnetic energy landscape and the difficulty of obtaining training data. Here we introduce a data-efficient neural network architecture to represent density functional theory total energy, atomic forces, and magnetic forces as functions of atomic and magnetic structures. Our approach incorporates the principle of equivariance under the three-dimensional Euclidean group into the neural network model. Through systematic experiments on various systems, including monolayer magnets, curved nanotube magnets, and moiré-twisted bilayer magnets of CrI3, we showcase the method's high efficiency and accuracy, as well as exceptional generalization ability. The work creates opportunities for exploring magnetic phenomena in large-scale materials systems. [ABSTRACT FROM AUTHOR]

Real-time play recognition and classification algorithms are crucial for automating video production and live broadcasts of sporting events. However, current methods relying on human pose estimation and deep neural networks introduce high latency on commodity hardware, limiting their usability in low-cost real-time applications. We present PlayNet, a novel approach to real-time handball play classification. Our method is based on Kalman embeddings, a new low-dimensional representation for game states that enables efficient operation on commodity hardware and customized camera layouts. Firstly, we leverage Kalman filtering to detect and track the main agents in the playing field, allowing us to represent them in a single normalized coordinate space. Secondly, we utilize a neural network trained in nonlinear dimensionality reduction through fuzzy topological data structure analysis. As a result, PlayNet achieves real-time play classification with under 55 ms of latency on commodity hardware, making it a promising addition to automated live broadcasting and game analysis pipelines. [ABSTRACT FROM AUTHOR]

This paper proposes a novel sensor scheduling scheme based on adaptive dynamic programming, which makes the sensor energy consumption and tracking error optimal over the system operational horizon for wireless sensor networks with solar energy harvesting. Neural network is used to model the solar energy harvesting. Kalman filter estimation technology is employed to predict the target location. A performance index function is established based on the energy consumption and tracking error. Critic network is developed to approximate the performance index function. The presented method is proven to be convergent. Numerical example shows the effectiveness of the proposed approach. [ABSTRACT FROM AUTHOR]

This paper proposes a new video-based face recognition method that uses video frames of a subject rotating his/her head. In the experiment discussed here, the manifolds of video frames embedded in a high-dimensional video space were extracted using neural network (NN)-based models. This increased the recognition rate in comparison with a simple NN architecture (from 72.9 to 81.3 %). These models were inspired by manifold interpretations of the brain's visual perception. Next, the pose and person manifolds were separated using the neurons trained in the hidden or bottleneck layer of the network. Finally, the separated manifolds were used to synthesize face images at different angles from a single frontal image. Using video frames to extract these manifolds produces higher-quality images than using manifolds extracted from single images. This improvement in image quality was verified using the structural similarity index measure. [ABSTRACT FROM AUTHOR]

Drought forecasting is a major component of a drought preparedness and mitigation plan. This paper focuses on an investigation of artificial neural networks (ANN) models for drought forecasting in the algerois basin in Algeria in comparison with traditional stochastic models (ARIMA and SARIMA models). A wavelet pre-processing of input data (wavelet neural networks WANN) was used to improve the accuracy of ANN models for drought forecasting. The standard precipitation index (SPI), at three time scales (SPI-3, SPI-6 and SPI-12), was used as drought quantifying parameter for its multiple advantages. A number of different ANN and WANN models for all SPI have been tested. Moreover, the performance of WANN models was investigated using several mother wavelets including Haar wavelet (db1) and 16 daubechies wavelets (dbn, n varying between 2 and 17). The forecast results of all models were compared using three performance measures (NSE, RMSE and MAE). A comparison has been done between observed data and predictions, the results of this study indicate that the coupled wavelet neural network (WANN) models were the best models for drought forecasting for all SPI time series and over lead times varying between 1 and 6 months. The structure of the model was simplified in the WANN models, which makes them very convenient and parsimonious. The final forecasting models can be utilized for drought early warning. [ABSTRACT FROM AUTHOR]

A multilayered neural network is a multi-input, multi-output nonlinear system in which network weights can be trained by using parameter estimation algorithms. In this paper, a novel training method is proposed. This method is based on the relatively new smooth variable structure filter (SVSF) and is formulated for feed-forward multilayer perceptron training. The SVSF is a state and parameter estimation that is based on the sliding mode concept and works in a predictor-corrector fashion. The SVSF training performance is tested on three benchmark pattern classification problems. Furthermore, a study is presented comparing the popular back-propagation method, the extended Kalman filter, and the SVSF. [ABSTRACT FROM AUTHOR]

Indoor radio propagation prediction modeling has been for a long time an important area of interest in research and development. In the literature, many propagation models are classified into empirical and deterministic models. The accuracy of both categories of models can be improved based on model calibration or tuning that uses real measurements collected in a given environment and frequency. Based on the availability of a huge measurement database, we aimed in this paper to develop a new propagation model using artificial neural networks. The new model is inspired from multi-wall one and will be available for the most used system bands, such as GSM, UMTS and WiFi. The model will be a multilayer perceptron and is trained with measured data using a back-propagation learning algorithm. Evaluated model performances show a high improvement in terms of accuracy compared to a calibrated multi-wall model. [ABSTRACT FROM AUTHOR]

Karst aquifers provide water resources for a large part of the Mediterranean population and water resources become a strategic problem during summer when the population increases due to tourism. To help managers to optimize the exploitation of karst aquifer waters this paper studies the ability of a neural model to efficiently simulate and forecast water table levels a few months ahead during the dry season. The neural model is a recurrent multilayer perceptron that learns the relations between inputs (mainly rainfall and pumping discharge) and output (water table level). After a training step using several years of data, the model is assessed on a never seen year to be validated. Particular attention is devoted to the dry season (May-September). The model achieved good forecast of the maximal observed drawdown. To investigate the usefulness of the model as an operational tool, various scenarios of future rainfalls and future pumping are proposed and associated forecasts are analysed. Scenarios are the following: zero-rainfall scenario and mean forecast rainfall 1 week ahead; in both cases, the model proposes accurate estimations of the maximal drawdown. In a second step, pumping scenarios are investigated: a mean pumping scenario during the dry season and the maximum pumping scenario taking into account the actual pumps. One more time, it is demonstrated that the maximal drawdown can be efficiently forecast for the dry season. The methodology is generic and can thus be used with profit by managers on other karst aquifers. [ABSTRACT FROM AUTHOR]

Because of numerous parameters existing in the Cloud's environment, it is helpful to introduce a general solution for dynamic resource provisioning in Cloud that is able to handle uncertainty. In this paper, a novel adaptive control approach is proposed which is based on continuous reinforcement learning and provides dynamic resource provisioning while dealing with uncertainty in the Cloud's environment. The proposed dynamic resource provisioner is a goal directed controller which provides ability of handling uncertainty specifically in Cloud's spot markets where competition between Cloud providers requires optimal policies for attracting and maintaining clients. This controller is aimed at hardly preventing from job rejection (as the primary goal) and minimizing the energy consumption (as the secondary goal). Although these two goals almost conflict (because job rejection is a common event in the process of energy consumption optimization), the results demonstrate the perfect ability of the proposed method with reducing job rejection down to near 0 % and minimizing energy consumption down to 9.55 %. [ABSTRACT FROM AUTHOR]

This paper presents a study to predict gender of individuals from scanned images of their handwritings. The proposed methodology is based on extracting a set of features from writing samples of male and female writers and training classifiers to learn to discriminate between the two. Writing attributes like slant, curvature, texture and legibility are estimated by computing local and global features. Classification is carried out using artificial neural networks and support vector machine. The proposed technique evaluated on two databases under a number of scenarios realized interesting results on predicting gender from handwriting. [ABSTRACT FROM AUTHOR]

The determination of weights is very important for neural network models. Nevertheless, the traditional feedforward neural networks usually use the method of random initial values to determine the weights of the neural networks, such a method would make the related neural network models possess unstable performance in accuracy. Therefore, in order to improve the prediction accuracy and efficiency of neural networks, a novel neural network model based on beetle antennae search (BAS) and weights and structure policy (WASP) is proposed and applied to the estimation of glomerular filtration rate (GFR). Through a series of numerical experiments, it is proved that the accuracy of the proposed neural network model for GFR estimation is higher than that of the model without BAS algorithm optimization. In addition, compared with the traditional one-step method, the method proposed in this article is more than 300% higher in the three performance indicators of MSE, MAE, and MAPE. [ABSTRACT FROM AUTHOR]

This paper focuses on the hybrid effects of Markovian switching, stochastic perturbation, and impulsive delay on neural networks. First, some novel generic criteria for Markovian switching neural networks with stochastic perturbation and impulsive delay are derived by establishing an extended Halanay differential inequality on impulsive dynamical systems. Second, our sufficient conditions ensuring synchronization are dependent on coupling and impulsive delay, and show coupling and impulsive effects on the synchronization of neural networks. Finally, simulation results demonstrate the effectiveness of the theoretical results. [ABSTRACT FROM AUTHOR]

In the optimization of artificial neural networks (ANNs) via evolutionary algorithms and the implementation of the necessary training for the objective function, there is often a trade-off between efficiency and flexibility. Pure software solutions on general-purpose processors tend to be slow because they do not take advantage of the inherent parallelism, whereas hardware realizations usually rely on optimizations that reduce the range of applicable network topologies, or they attempt to increase processing efficiency by means of low-precision data representation. This paper presents, first of all, a study that shows the need of heterogeneous platform (CPU-GPU-FPGA) to accelerate the optimization of ANNs using genetic algorithms and, secondly, an implementation of a platform based on embedded systems with hardware accelerators implemented in Field Pro-grammable Gate Array (FPGA). The implementation of the individuals on a remote low-cost Altera FPGA allowed us to obtain a 3x-4x acceleration compared with a 2.83 GHz Intel Xeon Quad-Core and 6x-7x compared with a 2.2 GHz AMD Opteron Quad-Core 2354. [ABSTRACT FROM AUTHOR]

In a paper on quantum neural networks, Zhou and Ding (Int. J. Theor. Phys. 46(12):3209-3215 ([])) proposed a new model of quantum perceptron denoted quantum M-P neural network and showed its functionality by an example. In this letter, we show that the proposed learning algorithm does not follow an unitary evolution and the proposed neuron can be efficiently simulated by a classical single layer neural network. [ABSTRACT FROM AUTHOR]

Feed-forward neural networks are commonly used for pattern classification. The classification accuracy of feed-forward neural networks depends on the configuration selected and the training process. Once the architecture of the network is decided, training algorithms, usually gradient descent techniques, are used to determine the connection weights of the feed-forward neural network. However, gradient descent techniques often get trapped in local optima of the search landscape. To address this issue, an ant colony optimization (ACO) algorithm is applied to train feed-forward neural networks for pattern classification in this paper. In addition, the ACO training algorithm is hybridized with gradient descent training. Both standalone and hybrid ACO training algorithms are evaluated on several benchmark pattern classification problems, and compared with other swarm intelligence, evolutionary and traditional training algorithms. The experimental results show the efficiency of the proposed ACO training algorithms for feed-forward neural networks for pattern classification. [ABSTRACT FROM AUTHOR]

This paper is concerned with the application of a Hybrid Algorithm (HA) to the determination of the Thermodynamic Activation Parameters (ATP) of a kinetic system of first order consecutive reactions. The 8 ATP's parameters involved in the Arrhenius and Eyring equations have been directly determined from the non-isothermal kinetic data without prior knowledge of the rate constants. AH is constituted by a combination of two algorithms based on different mathematical principles which are sequentially applied. In a first step, a ' soft modeling' method of Artificial Neural Networks (ANN) is applied and the obtained values of ATP's parameters are used as initial estimates of a new optimization algorithm (AGDC) applied in a second stage to improve the values of the final parameters. The great success of HA is the efficient resolution of the ambiguity of the results obtained by ANN. In addition, comparing with the classic algorithms, which present the known weak points, HA offers important advantages: (a) the lack of necessity to know a priori the initial estimates since they are calculated from ANN application, (b) the low probability of being trapped at local minima, saddle points, etc. by means of the exhaustive control and suitable correction of the movement vector during the optimization process, and (c) the simultaneous determination of a higher number of parameters endowed with very different orders of magnitude. [ABSTRACT FROM AUTHOR]

The paper presents two learning methods for nonlinear system identification. Both methods employ neural network models for representing state and output functions. The first method of learning nonlinear state space is based on using chaotic or noise signals in the training of state neural network so that the state neural network is designed to produce a sequence in a recursive way under the excitement of the system input. The second method of learning nonlinear state space has an observer neural network devoted to estimate the states as a function of the system inputs and the outputs of the output neural network. This observer neural network is trained to produce a state sequence when the output neural network is forced by the same sequence and then the state neural network is trained to produce the estimated states in a recursive way under the excitement of the system input. The developed identification methods are tested on a set of benchmark plants including a non-autonomous chaotic system, i.e. Duffing oscillator. Both proposed methods are observed much superior than well-known identification methods including nonlinear ARX, nonlinear ARMAX, Hammerstein, Wiener, Hammerstein-Wiener, Elman network, state space models with subspace and prediction error methods. [ABSTRACT FROM AUTHOR]

Maintaining the performance of reliable transport protocols, such as transmission control protocol (TCP), over wireless mesh networks (WMNs) is a challenging problem due to the unique characteristics of data transmission over WMNs. The unique characteristics include multi-hop communication over lossy and non-deterministic wireless mediums, data transmission in the absence of a base station, similar traffic patterns over neighboring mesh nodes, etc. One of the reasons for the poor performance of conventional TCP variants over WMNs is that the congestion control mechanisms in conventional TCP variants do not explicitly account for these unique characteristics. To address this problem, this paper proposes a novel artificial intelligence based congestion control technique for reliable data transfer over WMNs. The synergy with artificial intelligence is established by exploiting a carefully designed neural network (NN) in the congestion control mechanism. We analyze the proposed NN based congestion control technique in detail and incorporate it into TCP to create a new variant that we name as intelligent TCP or iTCP. We evaluate the performance of iTCP using both ns-2 simulations and real testbed experiments. Our evaluation results demonstrate that our proposed congestion control technique exhibits a significant improvement in total network throughput and average energy consumption per transmitted bit compared to the congestion control techniques used in other TCP variants. [ABSTRACT FROM AUTHOR]

Nonlinear system identification tends to provide highly accurate models these last decades; however, the user remains interested in finding a good balance between high-accuracy models and moderate complexity. In this paper, four balanced accuracy-complexity identification model families are proposed. These models are derived, by selecting different combinations of activation functions in a dedicated neural network design presented in our previous work (Romero-Ugalde et al. in Neurocomputing 101:170-180. doi:, ). The neural network, based on a recurrent three-layer architecture, helps to reduce the number of parameters of the model after the training phase without any loss of estimation accuracy. Even if this reduction is achieved by a convenient choice of the activation functions and the initial conditions of the synaptic weights, it nevertheless leads to a wide range of models among the most encountered in the literature. To validate the proposed approach, three different systems are identified: The first one corresponds to the unavoidable Wiener-Hammerstein system proposed in SYSID2009 as a benchmark; the second system is a flexible robot arm; and the third system corresponds to an acoustic duct. [ABSTRACT FROM AUTHOR]

In this paper, numerical techniques are developed for solving two-dimensional Bratu equations using different neural network models optimized with the sequential quadratic programming technique. The original two-dimensional problem is transformed into an equivalent singular, nonlinear boundary value problem of ordinary differential equations. Three neural network models are developed for the transformed problem based on unsupervised error using log-sigmoid, radial basis and tan-sigmoid functions. Optimal weights for each model are trained with the help of the sequential quadratic programming algorithm. Three test cases of the equation are solved using the proposed schemes. Statistical analysis based on a large number of independent runs is carried out to validate the models in terms of accuracy, convergence and computational complexity. [ABSTRACT FROM AUTHOR]

In this paper, stochastic techniques have been developed to solve the 2-dimensional Bratu equations with the help of feed-forward artificial neural networks, optimized with particle swarm optimization (PSO) and sequential quadratic programming (SQP) algorithms. A hybrid of the above two algorithms, referred to as the PSO-SQP method is also studied. The original 2-dimensional equations are solved by first transforming them into equivalent one-dimensional boundary value problems (BVPs). These are then modeled using neural networks. The optimization problem for training the weights of the network has been addressed using particle swarm techniques for global search, integrated with an SQP method for rapid local convergence. The methodology is evaluated by applying on three different test cases of BVPs for the Bratu equations. Monte Carlo simulations and extensive analyses are carried out to validate the accuracy, convergence and effectiveness of the schemes. A comparative study of proposed results is made with available exact solution, as well as, reported numerical results. [ABSTRACT FROM AUTHOR]

Adjusting control factors (independent variables) to achieve an optimal level of output (response variable) is usually required in many real-world manufacturing problems. Common optimization methods often begin with estimating the relationship between a response and independent variables. Among these techniques, response surface methodology (RSM), due to its simplicity, has recently attracted extensive attention. However, on the one hand, in some cases, the relationship between a response and independent variables is too complex to be estimated using polynomial regression models. On the other hand, solving the obtained optimization model is not easy by exact methods. This paper introduces a new methodology to solve multi-response statistical optimization problems. The novel hybrid approach of this research involves a modeling technique, a neural network methodology, and a genetic algorithm. The modeling technique that is selected among three common available methods is responsible to model the multi-response statistical problem. The neural network approach generates required input data, and finally, the genetic algorithm tries to optimize the model and find the adjusted levels of the control factors. At the end, the application and the performance of the proposed methodology are demonstrated using numerical examples. The results of several statistical analyses show that the proposed methodology is superior to other available methods in the literature. [ABSTRACT FROM AUTHOR]

Deep neural networks (DNNs) are vulnerable to adversarial examples that fool the models with tiny perturbations. Although adversarial attacks have achieved incredible attack success rates in the white-box setting, most existing adversaries often exhibit weak transferability in the black-box setting, especially for models with defense mechanisms. In this work, we reveal the cross-model channel redundancy and channel invariance of DNNs and thus propose two channel-augmented methods to improve the transferability of adversarial examples, namely, the channel transformation (CT) method and the channel-invariant Patch (CIP) method. Specifically, channel transformation shuffles and rewrites channels to enhance cross-model feature redundancy in convolution, and channel-invariant patches distinctly weaken different channels to achieve loss-preserving transformation. We compute the aggregated gradients of the transformed dataset to create adversaries with higher transferability. In addition, the two proposed methods can be naturally combined with each other and with almost all other gradient-based methods to further improve performance. Empirical results on the ImageNet dataset demonstrate that our attack methods exhibit higher transferability and achieve higher attack success rates than state-of-the-art gradient-based attacks. Specifically, our attack improves the average attack success rate from 86.9% to 91.0% on normally trained models and from 44.6% to 68.3% on adversarially trained models. [ABSTRACT FROM AUTHOR]

In this work, based on the signed graph theory, the model of quaternion-valued neural networks (QVNNs) with cooperative-competitive interaction is established. Based on the quaternion-valued signum function and some properties, a direct analytical method without separation is proposed and quaternion-valued control schemes are developed to investigate fixed-time bipartite synchronization (FXTBS) and prescribed-time bipartite synchronization (PTBS) of QVNNs. Lastly, simulation is performed to manifest the correctness of the theorems. [ABSTRACT FROM AUTHOR]