Best Set Of Weights Research Articles

A Cuckoo Search-Based Trained Artificial Neural Network for Symmetric Flow Problems

In this work, an artificial neural network based on the Cuckoo search algorithm (CS-ANN) is implemented for squeezing flow problems. Three problems are considered: the squeezing flow, the MHD squeezing flow, and the flow of the third-grade fluid past a moving belt. First, the approximation for the said nonlinear differential equations is explained and the proposed problems are transformed into the L2 norms of minimization problems. Then, a well-known Cuckoo search algorithm is used to minimize the norms of each problem to get the best set of weights for artificial neural networks. The outcome of the proposed method is displayed through graphs. Two cases for each problem are discussed consisting of the solution, error, weights, and fitness function, respectively. The numerical results for the state variables are displayed in Tables. The error analysis in each case proves the accuracy of our implemented technique. The results are validated through graphs by comparing CS-ANN results with the gradient descent method.

On Generalizing Sarle’s Bimodality Coefficient as a Path towards a Newly Composite Bimodality Coefficient

Determining whether a distribution is bimodal is of great interest for many applications. Several tests have been developed, but the only ones that can be run extremely fast, in constant time on any variable-size signal window, are based on Sarle’s bimodality coefficient. We propose in this paper a generalization of this coefficient, to prove its validity, and show how each coefficient can be computed in a fast manner, in constant time, for random regions pertaining to a large dataset. We present some of the caveats of these coefficients and potential ways to circumvent them. We also propose a composite bimodality coefficient obtained as a product of the weighted generalized coefficients. We determine the potential best set of weights to associate with our composite coefficient when using up to three generalized coefficients. Finally, we prove that the composite coefficient outperforms any individual generalized coefficient.

A Self-Adaptive Hybrid Bat Algorithm for Training Feedforward Neural Networks

Training feedforward neural network (FFNN) is a complex task in the supervised learning field. An FFNN trainer aims to find the best set of weights that minimizes classification error. This paper presents a new training method based on hybrid bat optimization with self-adaptive differential evolution to train the feedforward neural networks. The hybrid training algorithm combines bat and the self-adaptive differential evolution algorithm called BAT-SDE. BAT-SDE is used to better search in the solution space, which proves its effectiveness in large space solutions. The performance of the proposed approach was compared with eight evolutionary techniques and the standard momentum backpropagation and adaptive learning rate. The comparison was benchmarked and evaluated using seven bio-medical datasets and one large credit card fraud detection dataset. The results of the comparative study show that BAT-SDE outperformed other training methods in most datasets and can be an alternative to other training methods.

Survey on Heart Disease Anticipation Through Genetic Algorithm and BP-Neural Network

Medical Diagnosis Systems play a vital role in medical practice and are used by medical practitioners for diagnosis and treatment. In this paper, a medical diagnosis system is presented for predicting the risk of cardiovascular disease. This system is built by combining the relative advantages of genetic algorithm and neural network. Multilayered feed forward neural networks are particularly suited to complex classification problems. The weights of the neural network are determined using genetic algorithm because it finds acceptably good set of weights in less number of iterations. The dataset provided by University of California, Irvine (UCI) machine learning repository is used for training and testing. It consists of 303 instances of heart disease data each having 14 attributes including the class label. First, the dataset is preprocessed in order to make them suitable for training. Genetic based neural network is used for training the system. The final weights of the neural network are stored in the weight base and are used for predicting the risk of cardiovascular disease. The classification accuracy obtained using this approach is 94.17%.

A new approach for index construction: The case of the road user behavior index

In recent years, composite indicators have become increasingly recognized as a useful tool for performance evaluation, benchmarking, and decision-making by summarizing complex and multidimensional issues. In this study, we focus on the application of data envelopment analysis (DEA) on index construction in the context of road safety and highlight the shortcomings of using the classical DEA models. The DEA method assigns a weight to each indicator by selecting the best set of weights for the unit under evaluation. The flexibility in selecting the weights in the classical DEA approach may lead to two interrelated problems: compensability and unfairness. These shortcomings are, respectively, overcome traditionally by imposing weight restrictions and applying a common weights approach. However, the problem of evaluating a layered hierarchy of indicators with a common set of weights (CSW) has not been addressed in the literature. To fill this gap, we propose a new approach for index construction to determine an optimal CSW to assess all units simultaneously while reflecting the hierarchical structure of the indicators in the model. The applicability of the suggested common-weight approach is illustrated by a case study on constructing a road user behavior index for a set of European countries. From a theoretical point of view, our approach provides a fair and identical basis for evaluation and comparison of countries in terms of driver’s behaviors and, from a practical point of view, it significantly reduces the required computational burden for solving the formulated model. The obtained results clarify the sharper discrimination power of our model compared to the other methods in the literature.

Differential Neural Networks (DNN)

In this work, we propose an artificial neural network topology to estimate the derivative of a function. This topology is called a differential neural network because it allows the estimation of the derivative of any of the network outputs with respect to any of its inputs. The main advantage of a differential neural network is that it uses some of the weights of a multilayer neural network. Therefore, a differential neural network does not need to be trained. First, a multilayer neural network is trained to find the best set of weights that minimize an error function. Second, the weights of the trained network and its neuron activations are used to build a differential neural network. Consequently, a multilayer artificial neural can produce a specific output, and simultaneously, estimate the derivative of any of its outputs with respect to any of its inputs. Several computer simulations were carried out to validate the performance of the proposed method. The computer simulation results showed that differential neural networks are capable of estimating with good accuracy the derivative of a function. The method was developed for an artificial neural network with two layers; however, the method can be extended to more than two layers. Similarly, the analysis in this study is presented for two common activation functions. Nonetheless, other activation functions can be used as long as the derivative of the activation function can be computed.

A new weighted injury severity scoring system: better predictive power for adult trauma mortality

BackgroundAn accurate injury severity measurement is essential in the evaluation of trauma care and in outcome research. The traditional Injury Severity Score (ISS) does not consider the differential risks of the Abbreviated Injury Scale (AIS) from different body regions, and the three AIS involved in the calculation of ISS are given equal weights. The objective of this study was to develop a weighted injury severity scoring (wISS) system for adult trauma patients with better predictive power than the traditional Injury Severity Score (ISS).MethodsThe 2007–2014 National Trauma Data Bank (NTDB) Research Datasets were used. We identified adult trauma patients from the NTDB and then randomly split it into a study sample and a test sample. Based on the association between mortality and the Abbreviated Injury Scale (AIS) from each of the six ISS body regions in the study sample, we evaluated 12 different sets of weights for the component AIS scores used in the calculation of ISS and selected one best set of weights. Discrimination (areas under the receiver operating characteristic curve, sensitivity, specificity, positive predictive value, negative predictive value, concordance) and calibration were compared between the wISS and ISS.ResultsThe areas under the receiver operating characteristic curves from the wISS and ISS are all 0.83, and 0.76 vs. 0.73 for patients with ISS = 16–74 and 0.68 vs. 0.53 for patients with ISS = 25–74. The wISS showed higher specificity, positive predictive value, negative predictive value, and concordance when they were compared at similar levels of sensitivity. The wISS had better calibration than the ISS.ConclusionsBy weighting the AIS from different body regions, the wISS had significantly better predictive power for mortality than the ISS, especially in critically injured adults.

Optimizing a robust bi-objective supply chain network considering environmental aspects: a case study in plastic injection industry

ABSTRACTIn recent years, sustainability has been considered as a growing concern in supply chain network design (SCND). Due to legal requirements, economic profitability, environmental sensitivity, and awareness of customers, sustainable supply chain management has been important for government and industry over the years. The aims of this research are to propose a bi-objective mathematical programming model to integrate the flows of the entire chain to improve performance in conditions of uncertainty associated with the plastic injection industry. The proposed model seeks to minimize total costs including transportation costs, operating costs and fixed costs of setting up facilities and on the other hand, is looking to minimize the environmental emission to increase customer satisfaction. The model proposed based on two scenarios has been solved through the robust optimization approach. Also, to solve the proposed model, some of the multi-objective exact methods such as the weighted sum, weighted goal programming, and LP-metric method are used. Furthermore, a weight sensitivity analysis is applied to find the best set of weights for the objectives. Finally, the results are explained and suggestions for future research are presented.

An Adaptive Neural Network Controller Based on PSO and Gradient Descent Method for PMSM Speed Drive

In this paper, based on the combination of particle swarm optimization (PSO) algorithm and neural network (NN), a new adaptive speed control method for a permanent magnet synchronous motor (PMSM) is proposed. Firstly, PSO algorithm is adopted to get the best set of weights of neural network controller (NNC) for accelerating the convergent speed and preventing the problems of trapping in local minimum. Then, to achieve high-performance speed tracking despite of the existence of varying parameters in the control system, gradient descent method is used to adjust the NNC parameters. The stability of the proposed controller is analyzed and guaranteed from Lyapunov theorem. The robustness and good dynamic performance of the proposed adaptive neural network speed control scheme are verified through computer simulations.

Modelling minimum-time manoeuvering with global optimisation of local receding horizon control

In this paper, we explore the notion that a human driver uses a receding horizon model predictive control (MPC) scheme for minimum-time manoeuvering. However, MPC is an inherently sub-optimal control scheme because not all future information is incorporated into its finite preview horizon. In many practical applications, this sub-optimality is tolerated as the solution is sufficiently close to optimal. However, it is known that professional drivers have the ability to learn driving circuits and exploit its features to minimise their global manoeuvering time. In this paper, we will model their process with a cascaded optimisation structure. Therein, the inner-loop features a local MPC scheme tasked with finding the control inputs that achieve a blended objective of minimising time and maximising velocity in each preview horizon/distance. The outer loop of this cascaded structure computes the best set of weights for the two components of the local objectives in order to minimise the global manoeuvering time. The proposed cascaded optimisation and control approach is compared against a straight-forward fixed-cost time optimal MPC applied to minimum-time manoeuvering over two well-known race courses. The paper also includes an extended literature review and details of the computational formulation of the model approach.

Generic Content-Based Retrieval of Marker-Based Motion Capture Data.

In this work, we propose an original scheme for generic content-based retrieval of marker-based motion capture data. It works on motion capture data of arbitrary subject types and arbitrary marker attachment and labelling conventions. Specifically, we propose a novel motion signature to statistically describe both the high-level and the low-level morphological and kinematic characteristics of a motion capture sequence, and conduct the content-based retrieval by computing and ordering the motion signature distance between the query and every item in the database. The distance between two motion signatures is computed by a weighted sum of differences in separate features contained in them. For maximum retrieval performance, we propose a method to pre-learn an optimal set of weights for each type of motion in the database through biased discriminant analysis, and adaptively choose a good set of weights for any given query at the run time. Excellence of the proposed scheme is experimentally demonstrated on various data sets and performance metrics.

Theoretical insights into the augmented-neural-network approach for combinatorial optimization

The augmented-neural-network (AugNN) approach has been applied lately to some NP-Hard combinatorial problems, such as task scheduling, open-shop scheduling and resource-constraint project scheduling. In this approach the problem of search in the solution-space is transformed to a search in a weight-matrix space, much like in a neural-network approach. Some weight adjustment strategies are then used to converge to a good set of weights for a locally optimal solution. While empirical results have demonstrated the effectiveness of the AugNN approach vis-a-vis a few other metaheuristics, little theoretical insights exist which justify this approach and explain the effectiveness thereof. This paper provides some theoretical insights and justification for the AugNN approach through some basic theorems and also describes the algorithm and the formulation with the help of examples.

Case-based estimation of the risk of enterobiasis

To introduce an original case-based machine learning (ML) and prediction system Constud and its application on tabular data for estimation of the risk of enterobiasis among nursery school children in Estonia. The system consists of a software application and a knowledge base of observation data, parameters, and results. The data were obtained from anal swabs for the diagnosis of enterobiasis, from questionnaires for children's parents, observations in nursery schools and interviews with supervisors of the groups. The total number of studied children was 1905. Ten parallel ML processes were conducted to find the best set of weights for features and cases. The best goodness-of-fit according to the true skill statistic (TSS) was 0.381. Approximately equal fit can be reached using different sets of features. Cross-validation TSS of logit-regression and classification tree models was <0.24. In addition to the higher prediction fit, Constud is not sensitive to missing values of explanatory variables. The overall prevalence of enterobiasis was 22.8%; the mean of risk estimations was 47.8%. The overestimation of the prevalence in risk calculations can be interpreted as an inefficacy of the single swab analysis, or may be due to the relative constancy of the risk compared to the lability of infection and the applied objective function. In addition to the higher prediction fit, Constud is not sensitive to missing values of explanatory variables. The main risk factors of enterobiasis among nursery school children were the child's age, communication partners, habits, and cleanness of rooms in the nursery school. Mixed age groups at nursery schools also enhance the risk.

An algorithm for solving multiobjective problem with multiple Decision Makers

This article aims to develop an algorithm for solving multiobjective linear programming problems with multiple Decision Makers (DMs) which combines the interactive fuzzy method and a heuristic based on iterative computation. DMs' preference can be quantified with a membership function and a threshold which means the minimum level of acceptance of a solution. The proposed heuristic can automatically search for the best set of weights for the linear utility function. Since the global goal for the proposed heuristic is to minimise the average of all the membership value deviation from the threshold, over- as well as under-achievement in the objective function will be penalised. The algorithm is implemented with Matlab and GAMS and applied to a problem in warehouses with three different objective functions. The results illustrates that the developed algorithm is effective in providing efficient solutions with respect to different preference structures of DMs.

Optimal local weighted averaging methods in contour smoothing

In several applications where binary contours are used to represent and classify patterns, smoothing must be performed to attenuate noise and quantization error. This is often implemented with local weighted averaging of contour point coordinates, because of the simplicity, low-cost and effectiveness of such methods. Invoking the "optimality" of the Gaussian filter, many authors will use Gaussian-derived weights. But generally these filters are not optimal, and there has been little theoretical investigation of local weighted averaging methods per se. This paper focuses on the direct derivation of optimal local weighted averaging methods tailored towards specific computational goals such as the accurate estimation of contour point positions, tangent slopes, or deviation angles. A new and simple digitization noise model is proposed to derive the best set of weights for different window sizes, for each computational task. Estimates of the fraction of the noise actually removed by these optimum weights are also obtained. Finally, the applicability of these findings for arbitrary curvature is verified, by numerically investigating equivalent problems for digital circles of various radii.

Heuristic configuration of single hidden-layer feed-forward neural networks

For optimum statistical classification and generalization with single hidden-layer neural network models, two tasks must be performed: (a) learning the best set of weights for a network of k hidden units and (b) determining k, the best complexity fit. We contrast two approaches to construction of neural network classifiers: (a) standard back-propagation as applied to a series of single hidden-layer feed-forward nerual networks with differing number of hidden units and (b) a heuristic cascade-correlation approach that quickly and dynamically configures the hidden units in a network and learns the best set of weights for it. Four real-world applications are considered. On these examples, the back-propagation approach yielded somewhat better results, but with far greater computation times. The best complexity fit, k, for both approaches were quite similar. This suggests a hybrid approach to constructing single hidden-layer feed-forward neural network classifiers in which the number of hidden units is determined by cascade-correlation and the weights are learned by back-propagation.

Back propagation fails to separate where perceptrons succeed

It is widely believed that the back-propagation algorithm in neural networks, for tasks such as pattern classification, overcomes the limitations of the perceptron. The authors construct several counterexamples to this belief. They also construct linearly separable examples which have a unique minimum which fails to separate two families of vectors, and a simple example with four two-dimensional vectors in a single-layer network showing local minima with a large basin of attraction. Thus, back-propagation is guaranteed to fail in the first example, and likely to fail in the second example. It is shown that even multilayered (hidden-layer) networks can also fail in this way to classify linearly separable problems. Since the authors' examples are all linearly separable, the perceptron would correctly classify them. The results disprove the presumption, made in recent years, that, barring local minima, back-propagation will find the best set of weights for a given problem. >