Showing papers in "Neural Computing and Applications in 2013"

A survey of multi-view machine learning

Abstract: Multi-view learning or learning with multiple distinct feature sets is a rapidly growing direction in machine learning with well theoretical underpinnings and great practical success. This paper reviews theories developed to understand the properties and behaviors of multi-view learning and gives a taxonomy of approaches according to the machine learning mechanisms involved and the fashions in which multiple views are exploited. This survey aims to provide an insightful organization of current developments in the field of multi-view learning, identify their limitations, and give suggestions for further research. One feature of this survey is that we attempt to point out specific open problems which can hopefully be useful to promote the research of multi-view machine learning.

Bat algorithm for constrained optimization tasks

Abstract: In this study, we use a new metaheuristic optimization algorithm, called bat algorithm (BA), to solve constraint optimization tasks. BA is verified using several classical benchmark constraint problems. For further validation, BA is applied to three benchmark constraint engineering problems reported in the specialized literature. The performance of the bat algorithm is compared with various existing algorithms. The optimal solutions obtained by BA are found to be better than the best solutions provided by the existing methods. Finally, the unique search features used in BA are analyzed, and their implications for future research are discussed in detail.

Evaluation and prediction of blast-induced ground vibration at Shur River Dam, Iran, by artificial neural network

Abstract: The purpose of this article is to evaluate and predict blast-induced ground vibration at Shur River Dam in Iran using different empirical vibration predictors and artificial neural network (ANN) model. Ground vibration is a seismic wave that spreads out from the blasthole when explosive charge is detonated in a confined manner. Ground vibrations were recorded and monitored in and around the Shur River Dam, Iran, at different vulnerable and strategic locations. A total of 20 blast vibration records were monitored, out of which 16 data sets were used for training of the ANN model as well as determining site constants of various vibration predictors. The rest of the 4 blast vibration data sets were used for the validation and comparison of the result of ANN and different empirical predictors. Performances of the different predictor models were assessed using standard statistical evaluation criteria. Finally, it was found that the ANN model is more accurate as compared to the various empirical models available. As such, a high conformity (R 2 = 0.927) was observed between the measured and predicted peak particle velocity by the developed ANN model.

Decision tree classifiers for automated medical diagnosis

Abstract: Decision support systems help physicians and also play an important role in medical decision-making. They are based on different models, and the best of them are providing an explanation together with an accurate, reliable and quick response. This paper presents a decision support tool for the detection of breast cancer based on three types of decision tree classifiers. They are single decision tree (SDT), boosted decision tree (BDT) and decision tree forest (DTF). Decision tree classification provides a rapid and effective method of categorizing data sets. Decision-making is performed in two stages: training the classifiers with features from Wisconsin breast cancer data set, and then testing. The performance of the proposed structure is evaluated in terms of accuracy, sensitivity, specificity, confusion matrix and receiver operating characteristic (ROC) curves. The results showed that the overall accuracies of SDT and BDT in the training phase achieved 97.07 % with 429 correct classifications and 98.83 % with 437 correct classifications, respectively. BDT performed better than SDT for all performance indices than SDT. Value of ROC and Matthews correlation coefficient (MCC) for BDT in the training phase achieved 0.99971 and 0.9746, respectively, which was superior to SDT classifier. During validation phase, DTF achieved 97.51 %, which was superior to SDT (95.75 %) and BDT (97.07 %) classifiers. Value of ROC and MCC for DTF achieved 0.99382 and 0.9462, respectively. BDT showed the best performance in terms of sensitivity, and SDT was the best only considering speed.

Analysis of service satisfaction in web auction logistics service using a combination of Fruit fly optimization algorithm and general regression neural network

Abstract: When constructing classification and prediction models, most researchers used genetic algorithm, particle swarm optimization algorithm, or ant colony optimization algorithm to optimize parameters of artificial neural network models in their previous studies. In this paper, a brand new approach using Fruit fly optimization algorithm (FOA) is adopted to optimize artificial neural network model. First, we carried out principal component regression on the results data of a questionnaire survey on logistics quality and service satisfaction of online auction sellers to construct our logistics quality and service satisfaction detection model. Relevant principal components in the principal component regression analysis results were selected for independent variables, and overall satisfaction level toward auction sellers’ logistics service as indicated in the questionnaire survey was selected as a dependent variable for sample data of this study. In the end, FOA-optimized general regression neural network (FOAGRNN), PSO-optimized general regression neural network (PSOGRNN), and other data mining techniques for ordinary general regression neural network were used to construct a logistics quality and service satisfaction detection model. In the study, 4–6 principal components in principal component regression analysis were selected as independent variables of the model. Analysis results of the study show that of the four data mining techniques, FOA-optimized GRNN model has the best detection capacity.

A projective general linear group based algorithm for the construction of substitution box for block ciphers

Abstract: The substitution boxes are used in block ciphers with the purpose to induce confusion in data The design of a substitution box determines the confusion ability of the cipher; therefore, many different types of boxes have been proposed by various authors in literature In this paper, we present a novel method to design a new substitution box and compare its characteristics with some prevailing boxes used in cryptography The algorithm proposed in this paper apply the action of projective linear group PGL(2, GF(28)) on Galois field GF(28) The new substitution box corresponds to a particular type of linear fractional transformation (35z + 15)/(9z + 5) In order to test the strength of the proposed substitution box, we apply non-linearity test, bit independence criterion, linear approximation probability method, differential approximation probability method, strict avalanche criterion, and majority logic criterion This new technique to synthesize a substitution box offers a powerful algebraic complexity while keeping the software/hardware complexity within manageable parameters

Application of artificial neural networks for water quality prediction

Abstract: The term “water quality” is used to describe the condition of water, including its chemical, physical, and biological characteristics. Modeling water quality parameters is a very important aspect in the analysis of any aquatic systems. Prediction of surface water quality is required for proper management of the river basin so that adequate measure can be taken to keep pollution within permissible limits. Accurate prediction of future phenomena is the life blood of optimal water resources management. The artificial neural network is a new technique with a flexible mathematical structure that is capable of identifying complex non-linear relationships between input and output data when compared to other classical modeling techniques. Johor River Basin located in Johor state, Malaysia, which is significantly degrading due to human activities and development along the river. Accordingly, it is very important to implement and adopt a water quality prediction model that can provide a powerful tool to implement better water resource management. Several modeling methods have been applied in this research including: linear regression models (LRM), multilayer perceptron neural networks and radial basis function neural networks (RBF-NN). The results showed that the use of neural networks and more specifically RBF-NN models can describe the behavior of water quality parameters more accurately than linear regression models. In addition, we observed that the RBF finds a solution faster than the MLP and is the most accurate and most reliable tool in terms of processing large amounts of non-linear, non-parametric data.

Time series forecasting by evolving artificial neural networks with genetic algorithms, differential evolution and estimation of distribution algorithm

Abstract: Time series forecasting is an important tool to support both individual and organizational decisions (e.g. planning production resources). In recent years, a large literature has evolved on the use of evolutionary artificial neural networks (EANN) in many forecasting applications. Evolving neural networks are particularly appealing because of their ability to model an unspecified nonlinear relationship between time series variables. In this work, two new approaches of a previous system, automatic design of artificial neural networks (ADANN) applied to forecast time series, are tackled. In ADANN, the automatic process to design artificial neural networks was carried out by a genetic algorithm (GA). This paper evaluates three methods to evolve neural networks architectures, one carried out with genetic algorithm, a second one carried out with differential evolution algorithm (DE) and the last one using estimation of distribution algorithms (EDA). A comparative study among these three methods with a set of referenced time series will be shown. In this paper, we also compare ADANN forecasting ability against a forecasting tool called Forecast Pro® (FP) software, using five benchmark time series. The object of this study is to try to improve the final forecasting getting an accurate system.

Dynamic selection approaches for multiple classifier systems

Abstract: In this paper we propose a new approach for dynamic selection of ensembles of classifiers. Based on the concept named multistage organizations, the main objective of which is to define a multi-layer fusion function adapted to each recognition problem, we propose dynamic multistage organization (DMO), which defines the best multistage structure for each test sample. By extending Dos Santos et al.’s approach, we propose two implementations for DMO, namely DSA m and DSA c . While the former considers a set of dynamic selection functions to generalize a DMO structure, the latter considers contextual information, represented by the output profiles computed from the validation dataset, to conduct this task. The experimental evaluation, considering both small and large datasets, demonstrated that DSA c dominated DSA m on most problems, showing that the use of contextual information can reach better performance than other existing methods. In addition, the performance of DSA c can also be enhanced in incremental learning. However, the most important observation, supported by additional experiments, is that dynamic selection is generally preferred over static approaches when the recognition problem presents a high level of uncertainty.

A comparative study of generalized regression neural network approach and adaptive neuro-fuzzy inference systems for prediction of unconfined compressive strength of rocks

Abstract: The engineering properties of rocks play a significant role in planning and designing of mining and civil engineering projects. A laboratory database of mechanical and engineering properties of rocks is always required for site characterization and mineral exploitation. Due to discontinuous and variable nature of rock masses, it is difficult to obtain all physicomechanical properties of rocks precisely. Prediction of unconfined compressive strength from seismic wave velocities (Compressional wave, Shear wave) and density of rock using generalized regression neural network (GRNN) and adaptive neuro-fuzzy inference systems (ANFIS) can be appropriate and alternate methods to minimize the time and cost of tests. GRNN and ANFIS models were trained with 41 data sets using conjugate gradient descent algorithms and hybrid learning algorithm, respectively. Performance of both the models was examined with 15 testing data sets. In the present study, obtained network performance indices such as correlation coefficient, mean absolute percentage error, root mean square error and variance account for indicate high performance of predictive capability of GRNN system and closer to actual data over the ANFIS.

An ε-twin support vector machine for regression

Abstract: This study proposes a new regressor—e-twin support vector regression (e-TSVR) based on TSVR. e-TSVR determines a pair of e-insensitive proximal functions by solving two related SVM-type problems. Different form only empirical risk minimization is implemented in TSVR, the structural risk minimization principle is implemented by introducing the regularization term in primal problems of our e-TSVR, yielding the dual problems to be stable positive definite quadratic programming problems, so can improve the performance of regression. In addition, the successive overrelaxation technique is used to solve the optimization problems to speed up the training procedure. Experimental results for both artificial and real datasets show that, compared with the popular e-SVR, LS-SVR and TSVR, our e-TSVR has remarkable improvement of generalization performance with short training time.

A fast two-stage ACO algorithm for robotic path planning

Abstract: Ant colony optimization (ACO) algorithms are often used in robotic path planning; however, the algorithms have two inherent problems. On one hand, the distance elicitation function and transfer function are usually used to improve the ACO algorithms, whereas, the two indexes often fail to balance between algorithm efficiency and optimization effect; On the other hand, the algorithms are heavily affected by environmental complexity. Based on the scent pervasion principle, a fast two-stage ACO algorithm is proposed in this paper, which overcomes the inherent problems of traditional ACO algorithms. The basic idea is to split the heuristic search into two stages: preprocess stage and path planning stage. In the preprocess stage, the scent information is broadcasted to the whole map and then ants do path planning under the direction of scent information. The algorithm is tested in maps of various complexities and compared with different algorithms. The results show the good performance and convergence speed of the proposed algorithm, even the high grid resolution does not affect the quality of the path found.

Probabilistic neural network for breast cancer classification

Abstract: Among cancers, breast cancer causes second most number of deaths in women. To reduce the high number of unnecessary breast biopsies, several computer-aided diagnosis systems have been proposed in the last years. These systems help physicians in their decision to perform a breast biopsy on a suspicious lesion seen in a mammogram or to perform a short-term follow-up examination instead. In clinical diagnosis, the use of artificial intelligent techniques as neural networks has shown great potential in this field. In this paper, three classification algorithms, multi-layer perceptron (MLP), radial basis function (RBF) and probabilistic neural networks (PNN), are applied for the purpose of detection and classification of breast cancer. Decision making is performed in two stages: training the classifiers with features from Wisconsin Breast Cancer database and then testing. The performance of the proposed structure is evaluated in terms of sensitivity, specificity, accuracy and ROC. The results revealed that PNN was the best classifiers by achieving accuracy rates of 100 and 97.66 % in both training and testing phases, respectively. MLP was ranked as the second classifier and was capable of achieving 97.80 and 96.34 % classification accuracy for training and validation phases, respectively, using scaled conjugate gradient learning algorithm. However, RBF performed better than MLP in the training phase, and it has achieved the lowest accuracy in the validation phase.

A group theoretic approach to construct cryptographically strong substitution boxes

Abstract: In this paper, we present a method to construct a substitution box used in encryption applications. The proposed algorithm for the construction of substitution box relies on the linear fractional transform method. The design methodology is simple, while the confusion-creating ability of the new substitution box is complex. The strength of the proposed substitution box is evaluated, and an insight is provided to quantify the confusion-creating ability. In addition, tests are performed to assess the vulnerability of the encrypted data to algebraic and statistical attacks. The substitution box is critically analyzed by strict avalanche criterion, bit independent criterion, differential approximation probability test, linear approximation probability test, non-linearity test, and majority logic criterion. The performance of the proposed substitution box is also compared with those of some of the well-known counterparts including AES, APA, Gray, S8, Skipjack, Xyi, and prime of residue substitution boxes. It is apparent that the performance, in terms of confusion-creating ability, of the new substitution box is better than those of some of the existing non-linear components used in encryption systems. The majority logic criterion is applied to these substitution boxes to further evaluate the strength and usefulness.

A self-learning scheme for residential energy system control and management

Abstract: In this paper, we apply intelligent optimization method to the challenge of intelligent price-responsive management of residential energy use, with an emphasis on home battery use connected to the power grid. For this purpose, a self-learning scheme that can learn from the user demand and the environment is developed for the residential energy system control and management. The idea is built upon a self-learning architecture with only a single critic neural network instead of the action-critic dual network architecture of typical adaptive dynamic programming. The single critic design eliminates the iterative training loops between the action and the critic networks and greatly simplifies the training process. The advantage of the proposed control scheme is its ability to effectively improve the performance as it learns and gains more experience in real-time operations under uncertain changes of the environment. Therefore, the scheme has the adaptability to obtain the optimal control strategy for different users based on the demand and system configuration. Simulation results demonstrate that the proposed scheme can financially benefit the residential customers with the minimum electricity cost.

A multi-stage particle swarm for optimum design of truss structures

Abstract: The contribution of this study is to propose a multi-stage particle swarm optimization (MSPSO) for structural optimization. In this paper, three auxiliary improving mechanisms are added to the standard particle swarm optimization (PSO) in order to enhance its efficiency and reliability dealing with optimum design of truss structures. These mechanisms effectively accelerate the convergence rate of the PSO and also make it robust to attain better optimum solutions during various runs of the algorithm. The effectiveness of the MSPSO is illustrated by several benchmark structural optimization problems. Results demonstrate the efficiency and robustness of the proposed MSPSO algorithm compared to the standard version of the PSO.

Direct adaptive fuzzy backstepping control of uncertain nonlinear systems in the presence of input saturation

Abstract: In this paper, a novel direct adaptive fuzzy control approach is presented for uncertain nonlinear systems in the presence of input saturation. Fuzzy logic systems are directly used to tackle unknown nonlinear functions, and the adaptive fuzzy tracking controller is constructed by using the backstepping recursive design techniques. To overcome the problem of input saturation, a new auxiliary design system and Nussbaum gain functions are incorporated into the control scheme, respectively. It is proved that the proposed control approach can guarantee that all the signals of the resulting closed-loop system are semi-globally uniformly ultimately bounded (SGUUB), and the tracking error converges to a small neighborhood of the origin. A simulation example is included to illustrate the effectiveness of the proposed approach. Two key advantages of the scheme are that (i) the direct adaptive fuzzy control method is proposed for uncertain nonlinear system with input saturation by using Nussbaum function technique and (ii) The number of the online adaptive learning parameters is reduced.

Fast neural network learning algorithms for medical applications

Abstract: Measuring the blood urea nitrogen concentration is crucial to evaluate dialysis dose (Kt/V) in patients with renal failure. Although frequent measurement is needed to avoid inadequate dialysis efficiency, artificial intelligence can repeatedly perform the forecasting tasks and may be a satisfactory substitute for laboratory tests. Artificial neural networks represent a promising alternative to classical statistical and mathematical methods to solve multidimensional nonlinear problems. It also represents a promising forecasting application in nephrology. In this study, multilayer perceptron (MLP) neural network with fast learning algorithms is used for the accurate prediction of the post-dialysis blood urea concentration. The capabilities of eight different learning algorithms are studied, and their performances are compared. These algorithms are Levenberg–Marquardt, resilient backpropagation, scaled conjugate gradient, conjugate gradient with Powell–Beale restarts, Polak–Ribiere conjugate gradient and Fletcher–Reeves conjugate gradient algorithms, BFGS quasi-Newton, and one-step secant. The results indicated that BFGS quasi-Newton and Levenberg–Marquardt algorithm produced the best results. Levenberg–Marquardt algorithm outperformed clearly all the other algorithms in the verification phase and was a very robust algorithm in terms of mean absolute error (MAE), root mean square error (RMSE), Pearson’s correlation coefficient (\( R_{p}^{2} \)) and concordance coefficient (RC). The percentage of MAE and RMSE for Levenberg–Marquardt is 0.27 and 0.32 %, respectively, compared to 0.38 and 0.41 % for BFGS quasi-Newton and 0.44 and 0.48 % for resilient backpropagation. MLP-based systems can achieve satisfying results for predicting post-dialysis blood urea concentration and single-pool dialysis dose spKt/V without the need of a detailed description or formulation of the underlying process in contrast to most of the urea kinetic modeling techniques.

Adaptive optimal control for a class of continuous-time affine nonlinear systems with unknown internal dynamics

Abstract: This paper develops an online algorithm based on policy iteration for optimal control with infinite horizon cost for continuous-time nonlinear systems In the present method, a discounted value function is employed, which is considered to be a more general case for optimal control problems Meanwhile, without knowledge of the internal system dynamics, the algorithm can converge uniformly online to the optimal control, which is the solution of the modified Hamilton-Jacobi-Bellman equation By means of two neural networks, the algorithm is able to find suitable approximations of both the optimal control and the optimal cost The uniform convergence to the optimal control is shown, guaranteeing the stability of the nonlinear system A simulation example is provided to illustrate the effectiveness and applicability of the present approach

Prediction of power in solar stirling heat engine by using neural network based on hybrid genetic algorithm and particle swarm optimization

Abstract: In this paper, the model based on a feed-forward artificial neural network optimized by particle swarm optimization (HGAPSO) to estimate the power of the solar stirling heat engine is proposed. Particle swarm optimization is used to decide the initial weights of the neural network. The HGAPSO-ANN model is applied to predict the power of the solar stirling heat engine which data set reported in literature of china. The performance of the HGAPSO-ANN model is compared with experimental output data. The results demonstrate the effectiveness of the HGAPSO-ANN model.

Modular neural-SVM scheme for speech emotion recognition using ANOVA feature selection method

Abstract: The speech signal consists of linguistic information and also paralinguistic one such as emotion. The modern automatic speech recognition systems have achieved high performance in neutral style speech recognition, but they cannot maintain their high recognition rate for spontaneous speech. So, emotion recognition is an important step toward emotional speech recognition. The accuracy of an emotion recognition system is dependent on different factors such as the type and number of emotional states and selected features, and also the type of classifier. In this paper, a modular neural-support vector machine (SVM) classifier is proposed, and its performance in emotion recognition is compared to Gaussian mixture model, multi-layer perceptron neural network, and C5.0-based classifiers. The most efficient features are also selected by using the analysis of variations method. It is noted that the proposed modular scheme is achieved through a comparative study of different features and characteristics of an individual emotional state with the aim of improving the recognition performance. Empirical results show that even by discarding 22% of features, the average emotion recognition accuracy can be improved by 2.2%. Also, the proposed modular neural-SVM classifier improves the recognition accuracy at least by 8% as compared to the simulated monolithic classifiers.

Text categorization based on regularization extreme learning machine

Abstract: This article proposes a novel approach for text categorization based on a regularization extreme learning machine (RELM) in which its weights can be obtained analytically, and a bias-variance trade-off could be achieved by adding a regularization term into the linear system of single-hidden layer feedforward neural networks. To fit the input scale of RELM, the latent semantic analysis was used to represent text for dimensionality reduction. Moreover, a classification algorithm based on RELM was developed including the uni-label (i.e., a document can only be assigned to a unique category) and multi-label (i.e., a document can be assigned to multiple categories simultaneously) situations. The experimental results in two benchmarks show that the proposed method can produce good performance in most cases, and it could learn faster than popular methods such as feedforward neural networks or support vector machine.

Swallow swarm optimization algorithm: a new method to optimization

Abstract: This paper presents an exposition of a new method of swarm intelligence–based algorithm for optimization. Modeling swallow swarm movement and their other behavior, this optimization method represents a new optimization method. There are three kinds of particles in this method: explorer particles, aimless particles, and leader particles. Each particle has a personal feature but all of them have a central colony of flying. Each particle exhibits an intelligent behavior and, perpetually, explores its surroundings with an adaptive radius. The situations of neighbor particles, local leader, and public leader are considered, and a move is made then. Swallow swarm optimization algorithm has proved high efficiency, such as fast move in flat areas (areas that there is no hope to find food and, derivation is equal to zero), not getting stuck in local extremum points, high convergence speed, and intelligent participation in the different groups of particles. SSO algorithm has been tested by 19 benchmark functions. It achieved good results in multimodal, rotated and shifted functions. Results of this method have been compared to standard PSO, FSO algorithm, and ten different kinds of PSO.

The analysis of discrete artificial bee colony algorithm with neighborhood operator on traveling salesman problem

Abstract: The artificial bee colony (ABC) algorithm, inspired intelligent behaviors of real honey bee colonies, was introduced by Karaboga for numerical function optimization. The basic ABC has high performance and accuracy, if the solution space of the problem is continuous. But when the solution space of the problem is discrete, the basic ABC algorithm should be modified to solve this class optimization problem. In this study, we focused on analysis of discrete ABC with neighborhood operator for well-known traveling salesman problem and different discrete neighborhood operators are replaced with solution updating equations of the basic ABC. Experimental computations show that the promising results are obtained by the discrete version of the basic ABC and which neighborhood operator is better than the others. Also, the results obtained by discrete ABC were enriched with 2- and 3-opt heuristic approaches in order to increase quality of the solutions.

An online incremental learning support vector machine for large-scale data

Abstract: Support Vector Machines (SVMs) have gained outstanding generalization in many fields. However, standard SVM and most of modified SVMs are in essence batch learning, which make them unable to handle incremental learning or online learning well. Also, such SVMs are not able to handle large-scale data effectively because they are costly in terms of memory and computing consumption. In some situations, plenty of Support Vectors (SVs) are produced, which generally means a long testing time. In this paper, we propose an online incremental learning SVM for large data sets. The proposed method mainly consists of two components: the learning prototypes (LPs) and the learning Support Vectors (LSVs). LPs learn the prototypes and continuously adjust prototypes to the data concept. LSVs are to get a new SVM by combining learned prototypes with trained SVs. The proposed method has been compared with other popular SVM algorithms and experimental results demonstrate that the proposed algorithm is effective for incremental learning problems and large-scale problems.

Decentralized control of collaborative redundant manipulators with partial command coverage via locally connected recurrent neural networks

Abstract: This paper studies the decentralized control of multiple redundant manipulators for the cooperative task execution problem. Different from existing work with assumptions that all manipulators are accessible to the command signal, we propose in this paper a novel strategy capable of solving the problem even though there exists some manipulators unable to access the command signal directly. The cooperative task execution problem can be formulated as a constrained quadratic programming problem. We start analysis by re-designing the control law proposed in (Li et al. Neurocomputing, 2012), which solves the optimization problem recursively. By replacing the command signal with estimations with neighbor information, the control law becomes to work in the partial command coverage situation. However, the stability and optimality of the new system are not necessarily the same as the original system. We then prove in theory that the system indeed also globally stabilizes to the optimal solution of the constrained quadratic optimization problem. Simulations demonstrate the effectiveness of the proposed method.

Artificial neural networks applied to fetal monitoring in labour

Abstract: Birth asphyxia can result in death or permanent brain damage. To prevent it, the fetal heart rate (FHR) is recorded in labour on a paper strip. In clinical practice, the complicated FHR patterns are assessed by eye, which is error-prone, inconsistent and unreliable. Objective alternatives are needed and thus we investigated the applicability of feed-forward artificial neural networks (ANNs) for FHR analysis. Six FHR features were extracted and combined with six clinical parameters to form a feature space of 12 dimensions. The feature space was reduced to six dimensions by principal component analysis. Subsequently, a network committee of ten ANNs was trained with the data of 124 patients (a balanced set of 62 adverse, coded 1, and 62 normal outcomes, coded 0). The ANN committee was tested on another balanced set of 252 patients obtaining misclassification rate of 36%. Finally, the committee was tested on a large dataset of 7,568 patients (non-balanced). As the committee output continuously increased from 0 to 1, there was a consistent growth of the adverse outcome rate (from 0.26 to 5.3%) and the low umbilical pH rate (from 2.6 to 16.7%.) Based on this correlation between the committee output and the risk of compromise, we concluded that ANNs can be successfully applied to FHR monitoring in labour. However, extensive further work is necessary, for which we outline our plans. To our knowledge, this is the first time that an automated method for FHR diagnostic analysis has been tested on a database of this size.

Multiple classifier system for EEG signal classification with application to brain–computer interfaces

Abstract: In this paper, we demonstrate the use of a multiple classifier system for classification of electroencephalogram (EEG) signals. The main purpose of this paper is to apply several approaches to classify motor imageries originating from the brain in a more robust manner. For this study, dataset II from BCI competition III was used. To extract features from the brain signal, discrete wavelet transform decomposition was used. Then, several classic classifiers were implemented to be utilized in the multiple classifier system, which outperforms the reported results of other proposed methods on the dataset. Also, a variety of classifier combination methods along with genetic algorithm feature selection were evaluated and compared in order to diminish classification error. Our results suggest that an ensemble system can be employed to boost EEG classification accuracy.

Prediction of unconfined compressive strength of rock surrounding a roadway using artificial neural network

Abstract: The unconfined compressive strength (UCS) of rocks is an important design parameter in rock engineering and geotechnics, which is required and determined for rock mechanical studies in mining and civil projects. This parameter is usually determined through a laboratory UCS test. Since the preparation of high-quality samples is difficult, expensive and time consuming for laboratory tests, development of predictive models for determining the mechanical properties of rocks seems to be essential in rock engineering. In this study, an attempt was made to develop an artificial neural network (ANN) and multivariable regression analysis (MVRA) models in order to predict UCS of rock surrounding a roadway. For this, a database of laboratory tests was prepared, which includes rock type, Schmidt hardness, density and porosity as input parameters and UCS as output parameter. To make a database (including 93 datasets), different rock samples, ranging from weak to very strong types, are used. To compare the performance of developed models, determination coefficient (R 2), variance account for (VAF), mean absolute error (E a) and mean relative error (E r) indices between predicted and measured values were calculated. Based on this comparison, it was concluded that performance of the ANN model is considerably better than the MVRA model. Further, a sensitivity analysis shows that rock density and Schmidt hardness were recognized as the most effective parameters, whereas porosity was considered as the least effective input parameter on the ANN model output (UCS) in this study.

A framework for self-tuning optimization algorithm

Abstract: The performance of any algorithm will largely depend on the setting of its algorithm-dependent parameters. The optimal setting should allow the algorithm to achieve the best performance for solving a range of optimization problems. However, such parameter tuning itself is a tough optimization problem. In this paper, we present a framework for self-tuning algorithms so that an algorithm to be tuned can be used to tune the algorithm itself. Using the firefly algorithm as an example, we show that this framework works well. It is also found that different parameters may have different sensitivities and thus require different degrees of tuning. Parameters with high sensitivities require fine-tuning to achieve optimality.