This paper presents a Fourier-enhanced dynamic sequence-to-sequence latent graph neural network (Seq2SeqLatentGNN), a deep learning architecture for multi-node spatiotemporal forecasting in hydroelectric reservoir systems. The model integrates three key components: (i) a custom Fourier layer that analyzes global temporal patterns through frequency-domain transformations, (ii) a latent correlation graph convolutional network that infers relational structures between monitoring stations without requiring predefined adjacency matrices, and (iii) an attention-based sequence-to-sequence model that processes temporal dependencies while enabling multi-step forecasting. The architecture simultaneously learns graph structure and forecasting tasks, adapting to changing spatial relationships between reservoir nodes. The proposed architecture was evaluated using a comprehensive dataset derived from 19 interconnected hydroelectric reservoirs located in southern Brazil. The dataset encompasses multiple years of high-resolution (hourly) measurements, including reservoir water levels, inflow and outflow rates, precipitation records, and energy production metrics. Experimental results demonstrate that Seq2SeqLatentGNN achieves superior performance compared to conventional statistical models and contemporary machine learning methods, as measured by standard error metrics. Analysis of the learned latent correlations reveals meaningful spatial dependencies that align with hydrological principles. The model exhibits consistent performance across varying temporal patterns, adapts to regime transitions, and captures both periodic and nonstationary dynamics. The proposed architecture contributes to spatiotemporal forecasting by combining spectral processing, dynamic graph learning, and sequence modeling in a unified framework applicable to systems with evolving connectivity patterns. • Fourier layer captures global temporal patterns in reservoir systems. • Latent graph convolution uncovers dynamic spatial relationships. • Attention model processes temporal dependencies for multi-step forecasts. • Seq2SeqLatentGNN outperforms ML models on hydroelectric data.

Alzheimer's disease (AD), a neurodegenerative disorder, progressively impairs memory and cognitive functions.Magnetic resonance imaging (MRI) is used as AD diagnosis and progress monitoring method.Convolutional Neural Network (CNN) is a data-driven deep learning model containing layers transforming data input using convolution filters.The goal of this paper is to present an analysis of the CNN architectures for classifying AD diagnoses using functional brain MRI scans acquired by the experimental dataset from the Alzheimer's Disease Neuroimaging Initiative (ADNI).Results show CNNs variants such as InceptionV3 and In-ceptionResNetV2 as powerful computational tools for developing predictive neuroimaging biomarkers in AD diagnosis applications, with accuracy above 70%.

Random vector functional link networks (RVFLNs) simplify supervised learning to a purely linear issue that can be conducted by computing the Moore-Penrose pseudoinverse, eliminating the need to optimize the input to hidden layer weights and biases. This paper proposes a RVFLN approach enhanced through feature engineering pipeline using nonlinear transformations and Predictive Permutation Feature Selection method leveraging Markov blanket concepts combined with hyperparameter optimization using a tree-structured Parzen estimator. The optimized RVFLN hyperparameters include activation function type, number of nodes, regularization term, dropout rate, and feature inclusion/exclusion. This paper aims to enhance the accuracy of predictive modeling for carbon monoxide (CO) and nitrogen oxide (NOx) emissions in a gas turbine through the proposed RVFLN approach. Overall, the proposed RVFLN offered an effective tool for predicting gas turbine emissions considering the coefficient of determination (R2) for a 5-fold cross-validation scheme. For the CO emissions modeling, the RVFLN achieved the highest performance, with a test R2of 0.710 and a validation R2of 0.708. This was followed by the competitive performance of eXtreme Gradient Boosting and Random Forest (RF), which also demonstrated strong predictive capabilities, albeit slightly lower than RVFLN. For NOxemissions modeling, RVFLN delivered the best performance, achieving a test R2of 0.756 and a validation R2of 0.864. RF followed closely with a test R2of 0.729, while Ridge also demonstrated competitive results, attaining a test R2of 0.684 along with the shortest computation time. In contrast, LASSO (Least Absolute Shrinkage and Selection Operator) and ElasticNet underperformed significantly, with test R2values of 0.362 and 0.371, respectively.

This study proposes hybrid machine learning (ML) models to predict the daily urban water consumption scenario in a neighborhood Brazilian city. The framework evaluates various signal decomposition modes, including empirical wavelet transform (EWT), complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN), seasonal-trend decomposition (STL) using LOESS (LOcally Estimated Scatterplot Smoothing), and variational mode decomposition (VMD), to prepare the dataset. The decomposed data are combined with different ML models such as Bayesian regularized neural networks (BRNN), extreme learning machines (ELM), k-nearest neighbor (KNN), multilayer perceptron neural network (MLP), support vector regression with linear kernel function (SVRL), and support vector regression with radial basis function kernel (SVRR) for daily short- and long-term forecasting. The CEEMDAN-SVRL and VMD-SVRL hybrid models are found to have the best results in terms of statistical metrics and performance criteria, significantly improving the prediction accuracy and the stability of the results. The study demonstrates the potential of ML frameworks to improve water resource planning and management by accurately predicting water consumption scenarios. The results indicated that the VMD-SVRL model exhibited relatively better performance in most scenarios.

For the mechanical system to keep running as efficiently, safely, and effectively as feasible, the level-bearing vibration (LBV) in the car is required to be appropriately estimated. In light of recent regression progress, the broad learning system (BLS) models with mathematical proof of the universal approximation property are an alternative regression with origins in pseudo-inverse theory and compressed sensing. By utilizing a horizontal structure with feature nodes and enhancement nodes, BLS models provide options for deep learning to capture complex patterns. To address the limitations of single BLS in terms of generalization ability and robustness to noise, this study proposes an optimized ensemble of BLS models in bagging and stacking schemes for the LBV estimation dataset, measured at frequencies of 300 Hz and 1800 Hz. The LBV dataset is derived from simulations involving a lithium polymer cell model (ePLB C020) in an electric vehicle similar to the Nissan Leaf EV. The proposed BLS models are configured with 5-fold cross-validation scheme splitted into training (60%), validation (20%), and testing (20%) sets. Its hyperparameter tuning is based on a tree-structured Parzen estimator with an objective function given by minimization of the root mean square error (RMSE) of the validation set. Overall, the proposed ensemble BLS approaches offered an effective tool for LB estimation. The stacking BLS achieved the highest coefficient of determination (R2) train, validation, and test with 0.872, 0.821, and 0.815, respectively, followed by bagging BLS, eXtreme gradient boosting, and adaptive boosting. The results and comparison studies demonstrate ensemble BLS approaches exhibit better prediction accuracy and better fitting in data trends.

A comparison of proportional-integral-derivative (PID) controller tuning for a 2x2 Multiple-Input Multiple-Output (MIMO) water treatment plant model is presented. The PID tuning procedures are based on Multi-Objective Grey Wolf Optimizer (MOGWO) and Multi-Objective Grouped Grey Wolf Optimizer (MOGGWO) are performed under two different initial conditions. The tuning procedure is structured as a multi-objective optimization problem with four objective functions, the integral time-squared error is minimized in each output response, and the integral of the control actions is minimized in each input. MOGWO is a metaheuristic based on the hunting behavior of grey wolves, following the hierarchy and leadership procedures that these animals have, while the proposed MOGGWO is a new version proposed in this study with two different groups with different functionalities. The performance was analyzed by 50 runs based on statistical metrics. MOGGWO can find more local minima since its Pareto front is more dispersed than that found by MOGWO.

This work presents a performance comparison between the newly developed Manta Ray Foraging Optimization (MRFO) and two different variants created for tuning a decentralized fractional order proportional-integral-derivative (FOPID) controller for a multiple-input multiple-output (MIMO) application. The application consists of a ball mill pulverizing system to pulverize coal and maximize fuel efficiency. MRFO and its applications have the task of finding the optimal controller variable values to control the system's temperature and pressure. The MRFO is a metaheuristic based on the behavior of manta rays, and although it shows efficient performance it may be improved through the use of techniques which generate new variants of the algorithm. Two techniques were used in the de-velopment of new variants: Quantum mechanics and opposition-based learning. The three different versions of MRFO were used to minimize a custom fitness function which is a combination of the integral time squared error (ITSE) and the overshoot of the system response. Simulations were carried using Simulink and Matlab softwares. For analyzing the performances, statistical measures such as minimum, maximum, best, mean, median, and standard deviation of the fitness function over 50 runs were used. Additionally, the different variants were also compared for minimizing 10 benchmark functions. The results show that the use of the previously mentioned techniques improve the performance of the original MRFO in optimizing the FOPID to control the ball mill pulverizing system.

The wind energy forecast is an useful tool for wind farm production planning, and operation, facilitating decision making in terms of maintenance, electricity market clearing, and load sharing.This study proposes a cooperative ensemble learning model, using time series preprocessing, multi-objective optimization, and artificial intelligence to forecast wind energy generation in two wind farms in Brazil.Multi-objective optimization is employed to combine variational mode decompositionbased components of a model with bootstrap aggregation (bagging) and extreme learning machine models.Forecasting accuracy is evaluated through the root mean squared error, mean absolute error, mean absolute percentage error, and Diebold-Mariano hypothesis test.The empirical results suggest that proposed ensemble learning model achieved better forecasting performance than bootstrap stacking, machine learning, artificial neural networks, and statistical models, with values of approximately 12.76%, 25.25%, 31.91%, and 34.76%, respectively, in terms of root mean squared errors reduction for out-of-sample forecasting.

Forecasts can help in the decision-making process. Epidemiological forecasts are no different, they can help to evaluate the scenario and possible direction of disease spread, for guiding possible interventions. In this work, Echo State Networks (ESNs) are evaluated for COVID-19 (Coronavirus Disease 2019) cases and deaths forecasting ten days ahead. The chosen locations for the experiment are five states in Brazil, namely Sao Paulo (SP), Bahia (BA), Minas Gerais (MG), Rio de Janeiro (RJ), and Ceara (CE), the states with the most COVID-19 cases as of December 31, 2020. The results are evaluated using performance indexes RMSE (Root-mean-square error), MAE (Mean absolute error), and MAPE (Mean absolute percentage error). Results are compared with a common forecasting technique called ARIMA (Autoregressive Integrated Moving Average). The error signals are compared using Wilcoxon Signed-Rank Test, to evaluate the difference statistically. ESNs presented overall good results for a ten day horizon forecast regarding used performance metrics, but for the number of cases, ARIMA outperformed ESNs regarding RMSE, MAE, and MAPE in all but one state. For the number of deaths however, ESNs outperformed ARIMA in most states when the MAE is taken into account. ESNs are shown to be a solid forecasting model when compared with ARIMA, presenting comparable results and in some cases outperforming it.

Water consumption forecasting is essential for the development of efficient cities planning. Due to the non-linearities and relations of the water consumption with different factors the developing of an accurate forecasting system is challenging. This paper proposes a seasonal, trend and multiobjective ensemble learning model to forecast multi-step-ahead (one, two, and three-month-ahead) water consumption for two cities of Paraná state in Brazil. The proposed data analysis uses seasonal and trend decomposition using Loess (STL) to split the original data into the seasonal, trend, and residual components. In the next stage, the machine learning models named Support Vector Regression and Ridge Regression as well as the stochastic approach Gaussian Processes model are employed to train and predict the STL components. The previous components are weighted integrated to compose a heterogeneous ensemble learning of components obtaining the final forecasts. The elitist Non-Dominated Sorting Genetic Algorithm – version II (NSGA-II) is adopted to obtain the weights assigned to the components. The best model has better generalization out-of-sample considering the root mean squared error, mean absolute error, and mean absolute percentage error criteria in respect to minimization problem. Through developed comparisons, results showed that combining STL and multi-objective optimization with a heterogeneous ensemble learning can achieve high forecasting accuracy in comparison with some models. The framework proposed in this paper is effective to obtain reliable water consumption forecasting and can support and help future decision.

In multi-stage processes, decisions happen in an ordered sequence of stages. Many of them have the structure of dual funnel problem: as the sample size decreases from one stage to the other, the information increases. A related example is a selection process, where applicants apply for a position, prize or grant. In each stage, more applicants are evaluated and filtered out and from the remaining ones, more information is collected. In the last stage, decision-makers use all available information to make their final decision. To train a classifier for each stage becomes impracticable as they can underfit due to the low dimensionality in early stages or overfit due to the small sample size in the latter stages. In this work, we proposed a Multi-StaGe Transfer Learning (MSGTL) approach that uses knowledge from simple classifiers trained in early stages to improve the performance of classifiers in the latter stages. By transferring weights from simpler neural networks trained in larger datasets, we able to fine-tune more complex neural networks in the latter stages without overfitting due to the small sample size. We show that is possible to control the trade-off between conserving knowledge and fine-tuning using a simple probabilistic map. Experiments using real-world data show the efficacy of our approach as it outperforms other state-of-the-art methods for transfer learning and regularization.

In multi-stage processes, decisions occur in an ordered sequence of stages. Early stages usually have more observations with general information (easier/cheaper to collect), while later stages have fewer observations but more specific data. This situation can be represented as a dual funnel structure, in which the sample size decreases from one stage to the other while the information available about each instance increases. Training classifiers in this scenario is challenging since information in the early stages may not contain distinct patterns to learn (underfitting). In contrast, the small sample size in later stages can cause overfitting. We address both cases by introducing a framework that combines adversarial autoencoders (AAE), multitask learning (MTL), and multi-label semi-supervised learning (MLSSL). We improve the decoder of the AAE with MTL so it can jointly reconstruct the original input and use feature nets to predict the features for the next stages. We also introduce a sequence constraint in the output of an MLSSL classifier to guarantee the sequential pattern in the predictions. Using different domains (selection process, medical diagnosis), we show that our approach outperforms other state-of-the-art methods.

Alternative energy sources are becoming more and more common around the world. In order to reduce environmental pollution and CO2emissions, in addition to being an ideal solution to overcome the energy crisis. In this context, power energy stands out, as it is the most abundant and most widely available natural resource on the entire planet. Due to the high level of uncertainty of the factors that directly interfere in the generation of solar power, such as temperature and solar radiation, make predictions of solar power with high precision is a challenge. Thus, the objective of this article is to develop a forecasting model, through time series, that makes it possible to predict the production of power energy, using a database collected in a photovoltaic plant in Uruguay. For the development of the proposal, models (base-learners), pre-processing techniques and models (meta-learners) used in the Stacking-Ensemble Learnig (STACK) method were used, which were compared using the measurements of performance Relative Root Mean Square Error (RRMSE), Symmetric Mean Absolute Percentage Error (sMAPE) and Determination Coefficient (R2) in addition to statistical tests. In the end, it can be concluded that the combination Correlation Matrix (CORR) and Language Model (LM), from Layer-0 obtained the best results, in the three performance measures and the combination of models (base-learners) and pre-processing techniques (Layer-0) presented the best results when compared to Layer-1, obtaining satisfactory values in all performance measures.

We present a novel framework that can combine multi-domain learning (MDL), data imputation (DI) and multi-task learning (MTL) to improve performance for classification and regression tasks in different domains. The core of our method is an adversarial autoencoder that can: (1) learn to produce domain-invariant embeddings to reduce the difference between domains; (2) learn the data distribution for each domain and correctly perform data imputation on missing data. For MDL, we use the Maximum Mean Discrepancy (MMD) measure to align the domain distributions. For DI, we use an adversarial approach where a generator fill in information for missing data and a discriminator tries to distinguish between real and imputed values. Finally, using the universal feature representation in the embeddings, we train a classifier using MTL that given input from any domain, can predict labels for all domains. We demonstrate the superior performance of our approach compared to other state-of-art methods in three distinct settings, DG-DI in image recognition with unstructured data, MTL-DI in grade estimation with structured data and MDMTL-DI in a selection process using mixed data.

The development of accurate models to forecast electricity energy prices is a challenge due to the number of factors which can affect this commodity. In this paper, a hybrid multi-stage approach is proposed to forecast multi-stepahead (one, two and three-month-ahead) Brazilian commercial and residential electricity energy prices. The proposed data analysis combines the pre-processing named complementary ensemble empirical mode decomposition (CEEMD) in the first stage coupled with the coyote optimization algorithm (COA) to define the CEEMD's hyperparameters, aiming to deal with time series non-linearities and enhance the model's performance. On the next stage, four machine learning models named extreme learning machine, Gaussian process, gradient boosting machine, and relevance vector machine are employed to train and predict the CEEMD's components. Finally, in the final stage, the results of the previous step are directly integrated to compose a heterogeneous ensemble learning of components to obtain the final forecasts. In this case, a grid of models is obtained. The best model is one that has better generalization out-of-sample. Through developed comparisons, results showed that combining COA-CEEMD with a heterogeneous ensemble learning can develop accurate forecasts. The modeling developed in this paper is promising and can support decision making in electricity energy price forecasting.

Bitcoin is the leading currency in the cryptocurrency market capturing attention worldwide. Forecasting the Bitcoin price as accurate as possible is essential, but due to its high volatility this task is challenging. Many researchers try, through the years, to develop efficient models for predicting the Bitcoin price using several different data-driven approaches. The objective of this paper is to develop a novel decomposition-ensemble learning model that combines Variational Mode Decomposition (VMD) and Stacking-ensemble learning (STACK) with machine learning algorithms to forecast the Bitcoin price multi-step ahead. The algorithms are k-Nearest Neighbors, Support Vector Regression with Linear kernel, Feed-forward Artificial Neural Network with single-layer perceptron, Generalized Linear Model, and Cubist. Correlation matrix (CORR), principal component analysis (PCA), and Box-Cox transformation (BOXCOX) were used as data preprocessing techniques. Estimating the performance of the proposed models (namely VMD-STACK-CORR, VMD-STACK-PCA, and VMD-STACK-BOXCOX) using relative root mean square error, symmetric mean absolute percentage error, and absolute percentage error measures, defined that for one-day-ahead forecast VMD-STAK-BOXCOX model presented the better performance, and for two and three-days-ahead forecast VMD-STACK-CORR model was chosen, compared to VMD, STACK, and machine learning algorithms models' performance. Diebold-Mariano statistical test was conducted to evaluate a reduction in forecasting errors. Therefore, the proposed models (VMD-STACK-CORR, VMD-STACK-PCA, and VMD-STACK-BOXCOX) indeed forecast accurately Bitcoin price and outperformed the compared models (VMD, STACK, and machine learning models).

Artificial neural networks (ANN) has attracted attention of the academic community by the current progress that this technique has provided in speech recognition and digital media such as as image, video, audio, and signal processing. Some fields, as industrial process control and product development can be highly benefited by the development of techniques based on the proven potentialities of ANN models, allowing more accurate simulation, better adaptation to changing environments, and greater robustness in model-based fault diagnosis. Along with the advance of ANNs, there is a trend of open-source softwares use for soft computing which facilitates the access of the interested readers to implement their own codes and to explore other applications. Historically evolutionary algorithms such as the Genetic Algorithm (GA) have been implemented to evolve the architectures to search for solutions, in order to solve this fundamental issue that is still an open problem in the general case. Therefore, the present paper investigates the application of ANN to model the nonlinear aircraft turbojet engine through black-box approach. For that purpose it was used real-world measurements of aircraft engine's fuel and rotation as input and output, respectively. In order to facilitate the design, the ANN was optimized aiming to determine the best topology according to the one-step-ahead and free-run simulation. The results obtained encourage the use of automatically generated ANN architectures for dynamic system modeling.

The proportional-integral-derivative (PID) controller is widely adopted to control numerous process in industrial applications owing to its simplicity, clear functionality, and effectiveness. Due to a wide range of applications, distinct methods to determine PID gains are available in the specialized literature, and multiobjective optimization methods became an attractive approach to solve this problem, mainly when more complex systems are considered. Additionally, it is important to compare novel metaheuristics with traditional methods to follow the evolution of the state of the art. For that purpose, it was proposed a new multiobjective formulation of a PID-controlled antilock braking system (ABS) optimization through nondominated sorting genetic algorithm II (NSGA-II), multiobjective dragonfly algorithm (MODA), multiobjective salp swarm algorithm (MSSA) and a new version of MSSA with opposition based learning initialization and evolution. The experimental results were compared through spacing, euclidean distance and hypervolume metrics and showed that the change made in the MSSA (1) have improved the dominance and (2) the spread of Pareto front (PF) solutions.

Short-term forecasting plays an important role in the economic area. Several studies have been carried, where models with good forecast capacity, focusing on accuracy or stability, were built. Modeling only one of these characteristics without the other can lead to a model with lower generalization capacity. To deal with such situation, this study proposes an ensemble model (EM) to forecast one, two and three months ahead the 60 kg corn bag prices received by producers in the state of Parana (Brazil). Additionally, feature extraction by means of principal component analysis is employed. The EM is built using machine learning models as base (weak) learners (BL) combined by weighted sum. The adopted BL are: Extremely randomized trees, partial least squares, k-nearest neighbors, neural network, bagging and multivariate adaptive regression splines. The weights are chosen through multi-objective optimization, where bias and variance are minimized. The multi-objective differential evolution with spherical pruning algorithm is adopted, while physical programming is used in order to obtain the preferred set of weights. The model built is appointed as multi-objective ensemble model (MOEM). The performance of the model is evaluated using mean absolute percentage error, mean squared error and root mean squared error. Additionally, the Diebold-Mariano test is used to evaluate the reduction on forecasting errors. In general lines, the results show that forecasting using MOEM with two, three or four BL is more stable and accurate than forecasting with single BL. Therefore, this approach is recommended to make short-term forecast of corn prices, which leads to a more assertive decision making.

The behavior of natural phenomena has become one of the most popular sources for researchers to design optimization algorithms for scientific, computing and engineering fields. As a result, a lot of nature-inspired algorithms have been proposed in the last decades. Due to the numerous issues of the global optimization process, new algorithms are always welcome in this research field. This paper introduces the Coyote Optimization Algorithm (COA), which is a population based metaheuristic for optimization inspired on the canis latrans species. It contributes with a new algorithmic structure and mechanisms for balancing exploration and exploitation. A set of boundary constrained real parameter optimization benchmarks is tested and a comparative study with other nature-inspired metaheuristics is provided to investigate the performance of the COA. Numerical results and non-parametric statistical significance tests indicate that the COA is capable of locating promising solutions and it outperforms other metaheuristics on most tested functions.

Feature Selection (FS) is the process of identifying and separating relevant features of a dataset to obtain the best solutions for a pattern-classification or regression problem. The main benefits of FS include more accurate classification models, simplified interpretation of models and a reduction in the processing time required for classification. One of the main approaches used in FS involves wrappers. In this approach features are selected based on an evaluation performed by a classification algorithm. Recently, an optimization bio-inspired metaheuristic based on the intelligent behavior of crows called Crow Search Algorithm (CSA) has been proposed. CSA works based on the idea that crows store their excess food in hiding places and retrieve it when the food is needed. The main reasons of using CSA are its easy implementation, few control parameters to adjust, fast convergence speed and high efficiency. To further enhance the performance of the classical CSA algorithm, this paper proposes a new wrapper based in a “v-shaped” binarization of the CSA. The wrapper, which is referred to here as Binary CSA (BCSA), is applied to six benchmark data sets. The paper compares and discusses the advantages and disadvantages of the proposed technique in terms of classification accuracy, number of selected features and computational cost against some classical and state-of-art algorithms. The results were encouraging and showed that BCSA achieved very good results in terms of classification accuracy and also selected subsets with a small number of features with a relatively low computational cost.

Metaheuristics of the natural computing field have been proposed as an alternative to mathematical optimization approaches to address non convex problems involving large search spaces. In recent years a new optimization metaheuristic algorithm was proposed called Wind Driven Optimization (WDO). WDO is a stochastic nature-inspired paradigm based on atmospheric motion. In this paper, a modified version of WDO is proposed and evaluated, based on Lévy flights (or Lévy motions) to tune its control parameters, called Lévy WDO (LWDO). Lévy flight or anomalous diffusion process is a random walk characterized by Markov chain in which the step-lengths have a probability distribution that is heavy-tailed. To evaluate the multiobjective optimization performance of the WDO and the proposed LWDO, a benchmark for optimizing of a safety isolating transformer is adopted. In this paper, the transformer design optimization is treated as a multiobjective problem, with the aim to minimize both the total mass (iron and copper materials) and losses taking into consideration design constraints. Simulation results testify that the multiobjective LWDO is a promising approach for multiobjective optimization as it outperforms the WDO in multiobjective version and the classical NSGA-II (Non-dominated Sorting Genetic Algorithm, version II).

Time series forecasting plays a key role in many areas of science, finance and engineering, mainly for the estimation of trend or seasonality of a variable under observation, aiming to serve as basis for future purchase decisions, choice of design parameters or maintenance schedule. Artificial Neural Networks (ANNs) have proven to be suitable in linear or nonlinear functions mapping. However, the ANNs, implemented in its most simplistic form, tend to have a loss in overall performance. This work aims to obtain a prediction model for a short-term load problem through the usage of wavenets ensemble, which is an ANN approach capable in combining the best characteristics of each ensemble component, in order to achieve a higher overall performance. We adopted the usage of bootstrapping, cross-validation and the inputs decimation approaches for the ensemble construction. For the components selection, `constructive' and `no selection' methods were applied. Finally, the combination is held though simple average, mode or stacked generalization. The results show that it is possible to improve the generalization ability through effective committees depending on the methods used to construct the ensemble. The total relative improvement achieved in respect to the naive model, was over 95%, regardless the number of sub wavenets, and for the best component, the relative improvement was 93.91% using five wavenets. We conclude that the most frequent and effective set, but not always with the lower MSE (Mean Squared Error), was using constructive bagging with simple average.

Simulated annealing (SA) is a solo-search algorithm, trying to simulate the cooling process of molten metals through annealing to find the optimum solution in an optimization problem. SA selects a feasible starting solution, produces a new solution at the vicinity of it, and makes a decision by some rules to move to the new solution or not. However, the results found by SA depend on the selection of the starting point and the decisions SA makes. In this paper, in order to ameliorate the drawbacks of the algorithm, a population-based simulated annealing (PSA) algorithm is proposed. PSA uses the population's ability to seek different parts of the search space, thus hedging against bad luck in the initial solution or the decisions. A set of benchmark functions was used in order to evaluate the performance of PSA algorithm. Simulation results accentuate the superior capability of PSA in comparison with the other optimization algorithms.

Underwater image processing area has been considered an important topic within the last decades with important achievements. This kind of images are essentially characterized by their poor visibility because light is exponentially attenuated as it travels in the water and the scenes result poorly contrasted and hazy. On the other hand, image restoration takes into account the influence of the environment on the image in order to achieve an image with an improved quality. This technique consist of inverting the physical model of image formation. That model contains parameters which represent variables such as coefficients of absorption, scattering, among others. In this case, the quality of the restored image depends on the correct estimation of these parameters. In this work, an approach based on evolutionary optimization algorithms is proposed, for restoring underwater images by estimating the model parameters, and using two metrics for quality assessment. The degradation in the images has been simulated by using an image formation model. Results show that image restoration based on a Multi-Objective Differential Evolution (MODE) algorithm achieves images with good contrast and sharpness, being even better than the original image.

Proportional-Integral (PI) controllers remain as a practical and reliable solution for multivariable control for several industrial applications. Efforts to develop new tuning techniques fulfilling several performance indicators and guaranteeing robustness are worthwhile. Evolutionary multiobjective optimization (EMO) has been used for multivariable PI controller tuning, due to their flexibility and its advantages to depict the trade off among conflicting objectives. It is a regular practice bounding the search space as a hyperbox; nevertheless, the shape of the feasible space of PI parameters which are internally stable for a given control loop is irregular. Therefore, such hyperbox could enclose feasible and unfeasible solutions or contain a subset of the feasible set. In the former case, convergence capabilities of an algorithm could be compromised; in the latter case, search space is not fully explored. In this work, a coding mechanism is proposed in order to explore more efficiently the PI parameters feasible set (that is, all feasible solutions and only feasible solutions) in EMO. With the example provided, the advantages to approximate a Pareto front for 2, 3 and 5 objectives are shown, validating the mechanism as useful for EMO in multivariable PI controller tuning.

This paper presents a procedure for input selection and parameter estimation for system identification based on Radial Basis Functions Neural Networks (RBFNNs) models and Free Search Differential Evolution (FSDE). We adopt a cascaded evolutionary algorithm approach and problem decomposition to define the model orders and the related model parameters based on higher orders correlation functions. Thus, we adopt two distinct populations: the first to select the lags on the inputs and outputs of the system and the second to define the parameters for the RBFNN. We show the results when the proposed methodology is applied to model a coupled drives system with real acquired data. We use to this end the canonical binary genetic algorithm (selection of lags) and the recently proposed FSDE (definition of the model parameters), which is very convenient for the present problem for having few control parameters. The results show the validity of the approach when compared to a classical input selection algorithm.

In order to improve equipment efficiency in terms of performance, energy consumption and degradation for example, the industry has increased the use of control systems as the PID (proportional-integral-derivative) to a new baseline. This structure has few parameters to adjust and it is easy to implement practically. However, there are some requirements often included on multivariable systems that cannot be solved concurrently by classical methods. To solve this problem, the current paper approaches the application of Multiobjective Differential Evolution (MODE), Multiobjective Harmony Search (MOHS) and Multiobjective Particle Swarm Optimization (MOPSO) on multivariable PID controllers tuning. Moreover, an improved version of MOPSO (I-MOPSO) is proposed and its performance is compared with the other algorithms. In order to validate it under control systems, the optimization technique is applied on a two degree of freedom robotic manipulator. Finally, a detailed analysis is made on the I-MOPSO achievements.

Nature-inspired algorithms of the swarm intelligence field perform powerfully and efficiently in solving global optimization problems. Inspired by nature, these metaheuristic algorithms have obtained promising performance over continuous domains of optimization problems. Recently, a new swarm intelligence approach called firefly algorithm (FA) has emerged. The FA is a stochastic paradigm based on the idealized behavior of the flashing characteristics of fireflies. However, to achieve good performance with FA, the tuning of control parameters is essential as its performance is sensitive to the choice of the randomization parameter (α) setting. This paper introduces a FA approach combined with chaotic sequences generated by Zaslavskii map (FACZ) to tune the randomization parameter. Simulations of Loney's solenoid benchmark problem examine the effectiveness of the conventional FA and the proposed FACZ algorithms. Simulation results and comparisons with the FACZ demonstrated that the performance of the FA is promising in the Loney's solenoid case.

Differential evolution (DE) is an evolutionary algorithm (EA) that uses a rather greedy and less stochastic approach to solve optimization problems than other evolutionary methods [1]. Like other EAs, DE is a population-based, stochastic global optimizer, capable of working reliably in nonlinear and multimodal environments. Due to several features such as simplicity, efficiency and global search capabilities, DE rapidly became a successful paradigm of evolutionary computation. However, to achieve adequate performance with DE, the process of tuning the control parameters is essential as its performance is sensitive to the choice of both mutation and crossover settings. This paper proposes a DE algorithm with adaptive tuning of scaling factor (F), crossover rate (CR) and quasi-oppositional probability based on population's variance information - Adaptive Differential Evolution (ADE). Furthermore, ADE adopts a vector called Fm in each dimension of the optimization problem instead of single variable for F as presented in the classical DE approach. The proposed optimization method is validated on the test-bed proposed for the IEEE CEC'13 (IEEE Congress on Evolutionary Computation 2013) contest for real parameter single objective optimization with 28 benchmark functions. Simulation results over the benchmark functions demonstrate the effectiveness and usefulness of the proposed ADE method. This version of paper includes the ADE's performance on the 10, 30 and 50-dimensional benchmark functions.

The reliability-redundancy allocation problem can be approached as a mixed-integer programming problem. It has been solved by using optimization techniques such as dynamic programming, integer programming, and mixed-integer nonlinear programming. On the other hand, a broad class of meta-heuristics has been developed for reliability-redundancy optimization. Recently, a new meta-heuristics called firefly algorithm (FA) algorithm has emerged. The FA is a stochastic metaheuristic approach based on the idealized behavior of the flashing characteristics of fireflies. In FA, the flashing light can be formulated in such a way that it is associated with the objective function to be optimized, which makes it possible to formulate the firefly algorithm. This paper introduces a modified FA approach combined with chaotic sequences (FAC) applied to reliability-redundancy optimization. In this context, an example of mixed integer programming in reliability-redundancy design of an overspeed protection system for a gas turbine is evaluated. In this application domain, FAC was found to outperform the previously best-known solutions available.

Among the existing meta-heuristic optimization algorithms, a well-known branch is the differential evolution (DE). DE is a powerful population-based algorithm of evolutionary computation field designed for solving global optimization problems which only has a few control parameters. With an eye to improve the performance of DE, in this paper, a DE approach combined with a cultural algorithm technique based on normative knowledge (NDE) is investigated to estimate the heat transfer coefficient during freezing treatment by inverse analysis. Numerical results for inverse heat transfer problem demonstrate the applicability and efficiency of the NDE algorithm. In this application, NDE approach outperforms a classical DE approach in terms of quality of solution.

Synthesis of musical instruments or human voice is a time consuming process which requires theoretical and experimental knowledge about the synthesis engine. Commonly, performers need to deal with synthesizer interfaces and a process of trial and error for creating musical sounds similar to a target sound. This drawback can be overcome by adjusting automatically the synthesizer parameters using optimization algorithms. In this paper a hybrid particle swarm optimization (PSO) algorithm is proposed to solve the frequency modulation (FM) matching synthesis problem. The proposed algorithm takes advantage of a shuffle process for exchanging information between particles and applies the selective passive congregation and the opposition-based learning approaches to preserve swarm diversity. Both approaches for injecting diversity are based on simple operators, preserving the easy implementation philosophy of the particle swarm optimization. The proposed hybrid particle swarm optimization algorithm was validated for a three-nested FM synthesizer, which represents a 6-dimensional multimodal optimization problem with strong epistasis. Simulation results revealed that the proposed algorithm presented promising results in terms of quality of solutions.

The Harmony Search (HS) algorithm was originally conceptualized using the musical improvisation process of searching for a perfect state of harmony. The HS algorithm uses a random search, which is based on random selection, memory consideration, and pitch adjusting. This paper proposes a modified HS approach combined with differential evolution and chaotic sequences to solve the economic load dispatch problem of thermal generators with the valve-point effect. The proposed modified HS algorithm was validated in a power economic problem comprised by 10 generating units with valve-point effects and multiple fuels for a load demand of 2500 MW. Simulation results and performance analysis show that the modified HS algorithm presented promising results when compared with results of other optimization methods reported in recent literature.

In this paper a Modified Differential Evolution (MDE) is proposed and its performance for solving the inverse heat transfer problem is compared with Genetic Algorithm with Floating-point representation (GAF) and classical Differential Evolution (DE). The inverse analysis of heat transfer has some practical applications, for example, the estimation of radioactive and thermal properties, such as the conductivity of material with and without the temperatures dependence of diffusive processes. The inverse problems are usually formulated as optimization problems and the main objective becomes the minimization of a cost function. MDE adapts a concept originally proposed in particle swarm optimization design for the dynamic adaptation of mutation factor. Using a piecewise function for apparent thermal conductivity as a function of the temperature data, the heat transfer equation is able to estimate the unknown variables of the inverse problem. The variables that provide the beast least squares fit between the experimental and predicted time-temperatures curves were obtained. Numerical results for inverse heat transfer problem demonstrated the applicability and efficiency of the MDE algorithm. In this application, MDE approach outperforms the GAF and DE best solutions.

A fundamental problem in robotics consists in trajectory planning. The main task of path planning for robot manipulators is to find an optimal collision-free trajectory from an initial to a final configuration. Furthermore, trajectory planning is devoted to generate the reference inputs for the control system of the manipulator, so as to be able to execute the motion. Many important contributions to this problem have been made in recent years. Recently, techniques based on metaheuristics of natural computing, mainly evolutionary algorithms (EA), have been successfully applied to a large number of robotic applications, including the generation of optimized trajectories for robot manipulators. The aim of this paper is to evaluate a modified Biogeography-based Optimization (BBO) approach based on Predator-Prey concepts (PPBBO) to solve the trajectory planning of a robot manipulator. Simulation experiments are carried on a robot manipulator with three degrees of freedom (3-DOF) to illustrate the efficacy of the BBO approach. Biogeography deals with the geographical distribution of biological organisms. BBO is an optimization method which is motivated by the nature's way of distributing habitats. Similar to genetic algorithms, BBO is a population-based stochastic global optimizer. However, in BBO, problem solutions are represented as islands, and the sharing of features between solutions is represented as migration between islands. Results demonstrated that the proposed PPBBO approach converged to promising solutions in terms of quality and convergence rate when compared with the classical BBO.

In this article, we propose a nonlinear forecasting model based on radial basis function neural networks (RBF-NNs) with Gaussian activation functions and robust clustering algorithms to model the conditional mean and a parametric generalized autoregressive conditional heteroskedasticity (GARCH) specification to model the conditional volatility. Instead of calibrating the parameters of the RBF-NNs via numerical simulations, we propose a novel estimation procedure by which the number of basis functions, their corresponding widths and the parameters of the GARCH model are jointly estimated via maximum likelihood along with a genetic algorithm to maximize the likelihood function. We use this model to provide hour-ahead point and direction-of-change forecasts of the Spanish electricity pool prices.

The particle swarm optimization (PSO) algorithm is a member of the wide category of swarm intelligence methods for solving global optimization problems. Its basic idea is the simulation of simplified animal social behaviors such as fish schooling and bird flocking. PSO algorithms are attracting attentions in recent years, due to their ability of keeping good balance between convergence and diversity maintenance. Several attempts have been made to improve the performance of the original PSO algorithm. In this paper, a modified version of the original PSO based on Cauchy distribution and dynamic adaptation of inertia factor, named modified PSO (MPSO), is proposed. to estimate the unknown variables of an inverse heat transfer problem. To validate the optimization performance of the proposed MPSO, an inverse heat transfer problem is illustrated and the algorithm has to estimate its unknown variables. The results testify that the MPSO can perform well in an inverse heat transfer problem.

High computational cost for solving large engineering optimization problems point out the design of parallel optimization algorithms. Population based optimization algorithms provide parallel capabilities that can be explored by their implementations done directly in hardware. This paper presents a hardware implementation of Particle Swarm Optimization algorithms using an efficient floating-point arithmetic which performs the computations with high precision. All the architectures are parameterizable by bit-width, allowing the designer to choose the suitable format according to the requirements of the optimization problem. Synthesis and simulation results demonstrate that the proposed architecture achieves satisfactory results obtaining a better performance in therms of elapsed time than conventional software implementations.

The reliability-redundancy allocation problem can be approached as a mixed-integer programming problem. It has been solved by using optimization techniques such as dynamic programming, integer programming, and mixed-integer nonlinear programming. On the other hand, a broad class of meta-heuristics has been developed for reliability-redundancy optimization. Recently, a new meta-heuristics called harmony search (HS) algorithm has emerged. HS was conceptualized using an analogy with music improvisation process where music players improvise the pitches of their instruments to obtain better harmony. This paper introduces a modified HS approach combined with an operator of differential evolution — a paradigm of evolutionary computation — to solve optimization problems in reliability engineering. In this context, an example of mixed integer programming in reliability-redundancy design of an over-speed protection system for a gas turbine is evaluated. In this application domain, HS was found to outperform the previously best-known solutions available.

A new class of meta-heuristics called SOMA (Self-Organizing Migrating Algorithm) was proposed in recent literature. SOMA works on a population of potential solutions called specimen and it is based on the self-organizing behavior of groups of individuals in a “social environment”. This paper proposes a modified SOMA approach to solving the economic load dispatch problem of thermal generators with the valve-point effect. To show the performance of the proposed modified SOMA algorithm based on fundamentals of normative knowledge in cultural algorithms, which was applied to test the power economic problem comprised 10 generating units with valve-point effects and multiple fuels for the load demands of 2400 MW. Simulation results show that the classical and modified SOMA algorithms are efficient and have good convergence property when compared with results of other optimization methods reported in the literature.

The reliability-redundancy allocation problem is a mixed-integer programming problem. It has been solved by using optimization techniques such as dynamic programming, integer programming, mixed-integer non-linear programming, heuristics, and meta-heuristics. Meanwhile, the development of meta-heuristics has been an active research area in optimizing system reliability wherein the redundancy, the component reliability, or both are to be determined. In recent years, a broad class of stochastic algorithms, such as simulated annealing, evolutionary computation, and swarm intelligence algorithms, has been developed for reliability-redundancy optimization of systems. Recently, a new class of stochastic optimization algorithm called SOMA (Self-Organizing Migrating Algorithm) has emerged. SOMA works on a population of potential solutions called specimen, and is based on the self-organizing behavior of groups of individuals in a “social environment”. This paper introduces a modified SOMA approach based on a Gaussian operator to solve reliability-redundancy optimization problems. In this context, three examples of mixed integer programming in reliability-redundancy design problems are evaluated. In this application domain, SOMA was found to outperform the previously best-known solutions available.

Differential evolution (DE) is a powerful population-based algorithm of evolutionary computation field designed for solving global optimization problems. The potentialities of DE are its simple structure, easy use, convergence speed and robustness. However, the control parameters and learning strategies involved in DE are highly dependent on the problems under consideration. Choosing suitable parameter values requires also previous experience of the user. Despite its crucial importance, there is no consistent methodology for determining the control parameters of a DE. In this paper, different DE approaches combined with a cultural algorithm technique based on normative and situational knowledge are proposed as alternative methods to solving the economic load dispatch problem of thermal generators with valve-point effect. The DE approaches are validated for a test system consisting of 13 thermal generators whose nonsmooth fuel cost function takes into account the valve-point loading effects. Numerical results indicate that performance of the cultural DE present best results when compared with previous optimization approaches in solving load dispatch problems with the valve-point effect.

This paper introduces an approach to deal with IMS (Intelligent Manufacturing System) models, in order to help the development of Decision Support Systems, Simulators and Performance Evaluators. The complexity of such systems is growing continuously requiring knowledge and information exchange between different manufacturing functions with more and more efficiency. This paper explores some potentialities of soft computing approaches and intelligent agent-oriented design including artificial neural networks, fuzzy systems, and evolutionary computation. The rationale is in order to support decision making processes performed by the autonomous and co-operative units of an IMS.

During the history of science of computational intelligence, many evolutionary algorithms approaches were proposed having more or less success in solving various optimization problems. In this context, the Particle Swarm Optimization (PSO) is a bio-inspired optimization mechanism based on the metaphor of social behaviour of birds flocking and fish schooling in search for food. Inspired by the classical PSO method and quantum mechanics theories, this work presents a quantum-behaved PSO (QPSO) approach using Gaussian probability distribution function (G-QPSO). Numerical simulations based on optimized proportional-integral-derivative (PID) control of an automatic regulator voltage system for nominal system parameters and step reference voltage input demonstrate the effectiveness and efficiency of G-QPSO approach. Simulation results of G-QPSO to determine the PID parameters are compared with the classical PSO and QPSO.

The identification of a nonlinear system with chaotic behavior denoted Chua's oscillator by using fuzzy models intertwined with particle swarm optimization (PSO) method is presented. This hybrid approach is applied to experimental data generated by an inductorless Chua's circuit that consists of an electronic chaotic oscillator. Chua's circuit has been used as a test platform by various scientific and engineering communities related to the study of chaos due its great potencial in technological applications, for instance, telecommunications, cryptography and physics. Fuzzy set theory has been evolved as a powerful modeling tool that can cope with uncertainties and nonlinearities in modeling and identification procedures. The identification of a optimized fuzzy model Takagi-Sugeno (T-S) fuzzy model involves two primary tasks: parameter tuning and structure optimization. The premise part of production rules is optimized here by using the particle swarm optimization method. In turn, least mean squares technique is applied to the consequent part of a T-S fuzzy model. Results indicate PSO method and Least Mean Square technique succeeded in constructing a T-S fuzzy model when dealing with chaotic dynamics obtained through experimental data supplied by inductorless Chua's electronic circuit.

Particle swarm optimization (PSO) is a population-based stochastic optimization technique, originally developed by Eberhart and Kennedy, inspired by simulation of a social psychological metaphor instead of the survival of the fittest individual. In PSO, the system (swarm) is initialized with a population of random solutions (particles) and searches for optima using cognitive and social factors by updating generations. PSO has been successfully applied to a wide range of applications, mainly in solving continuous nonlinear optimization problems. Based on the PSO and chaos theories, this paper discusses the use of a chaotic PSO approach hybridized with an implicit filtering (IF) technique to optimize performance of economic dispatch problems. The chaotic PSO with chaos sequences is the global optimizer and the IF is used to fine-tune the chaotic PSO run in sequential manner. The hybrid methodology is validated for a test system consisting of 13 thermal units whose incremental fuel cost function takes into account the valve-point loading effects.

Nowadays the high degree of the industrial activity as well as the increasing society life standard have been accompanied by a growing waste generation which represents one of the most serious environmental problems. The possibility use of some industrial wastes in the cement production, as an alternative source of secondary raw materials, as well as alternative secondary fuels have been a viable path to reduce the cement industries production cost. The main concerns about the use of these fuels are the effects in the cement performance and the environmental impacts that they can cause. Through an optimization model the influence of these fuels in the cement Portland properties are analyzed. In this model the Sequential Quadratic Programming (SQP), Genetic Algorithm (GA) and Differential Evolution (DE) are applied taking account the raw material, fuels cost, clinker quality and the environmental impact, such as the consumption of the energy requested in the grinding for the cement production.

Particle swarm optimization (PSO) is a population-based swarm intelligence algorithm driven by the simulation of a social psychological metaphor instead of the survival of the fittest individual. Based on the swarm intelligence theory, this paper discusses the use of PSO approaches using an operator and based on the Gaussian probability distribution function as a population space of a cultural algorithm, called cultural Gaussian PSO (GPSO-CA). Cultural algorithms are mechanisms that incorporate domain knowledge obtained during the evolutionary process, which increase the efficiency of the search process. These approaches are employed in a well-studied continuous optimization problem of mechanical engineering design.

Studies with the Gaussian and Cauchy probability distributions have shown that the performance of the standard PSO algorithm can be improved. But these versions may also get stuck in local minima when optimizing functions with many local minima in high dimensional search space. In this paper, we will provide new results with PSO using the Exponential probability distribution aiming at improvement in performance. This version of the algorithm, termed PSO-E, was tested on a suite of well-known benchmark functions with many local optima and the results were compared with those obtained by the standard PSO (constriction factor). Simulation results show the suitability of PSO– E.

A method of nonlinear identification based on the Takagi-Sugeno (TS) fuzzy model and optimization procedure is proposed in this paper. New chaotic particle swarm optimization algorithms based on Zaslavskii chaotic map sequences combined with efficient Gustafson-Kessel (GK) clustering algorithm are proposed here for the design of the premise part of production rules, while the least mean squares technique is utilized for the subsequent part of the production rules of a TS fuzzy model. The numerical results presented here indicate that the particle swarm optimization (PSO) and particularly the chaotic PSO combined with GK algorithms are effective in building a good TS fuzzy model for nonlinear identification of a nonlinear yo-yo motion control system.

Artificial neural networks, in particular, feedforward multilayer networks and basis function networks, have gradually established themselves as a usual tool in approximating complex nonlinear systems. B-spline networks, a type of basis function neural network, are normally trained by gradient-based methods, which may fall into local minima during the learning phase. In order to overcome the drawbacks encountered by conventional learning methods, particle swarm optimization - a swarm intelligence methodology - can provide a stochastic global search of B-spline networks for nonlinear system identification. In this paper, a modified particle swarm optimization algorithm using Gaussian and Cauchy probability distributions are applied to adjust the control points of B-spline neural networks. Simulation results for the identification of Rössler systems are provided and demonstrate the effectiveness and robustness of the proposed identification scheme.

Artificial neural networks and fuzzy systems, have gradually established themselves as popular tools in approximating complicated nonlinear systems and time series forecasting. This paper investigates the hypothesis that the nonlinear mathematical models of multilayer perceptron and radial basis function neural networks and the Takagi-Sugeno (TS) fuzzy system are able to provide a more accurate out-of-sample forecast than the traditional AutoRegressive Moving Average (ARMA) and ARMA Generalized AutoRegressive Conditional Heteroskedasticity (ARMA-GARCH) linear models. Using series of Brazilian exchange rate (R$/US$) returns with 15 min., 60 min., 120 min., daily and weekly basis, the out-pf-sample one-step-ahead forecast performance is compared. Results indicate that forecast performance is strongly related to the series' frequency and the forecasting evaluation shows that nonlinear models perform better than their linear counterparts. In the trade strategy based on forecasts, nonlinear models achieve higher returns when compared to a buy-and-hold strategy and to the linear models.

This paper describes the application of differential evolution approaches to the optimization of a supply chain. Although simplified, this supply chain included stocks, production, transportation and distribution, in an integrated production-inventory-distribution system. The supply chain problem model is presented as well as a short introduction to each evolutionary algorithm. Differential evolution (DE) is an emergent evolutionary algorithm that offers three major advantages: it finds the global minimum regardless of the initial parameter values, it involves fast convergence, and it uses few control parameters. Inspired by the chaos theory, this work presents a new global optimization algorithm based on different DE approaches combined with chaotic sequences (DEC), called chaotic differential evolution algorithm. The performance of three evolutionary algorithm approaches (genetic algorithm, DE and DEC) and branch and bound method were evaluated with numerical simulations. Results were also compared with other similar approach in the literature. DEC was the algorithm that led to better results, outperforming previously published solutions. The simplicity and robustness of evolutionary algorithms in general, and the efficiency of DEC, in particular, suggest their great utility for the supply chain optimization problem, as well as other logistics-related problems.

Particle swarm optimization (PSO) is a population-based swarm intelligence algorithm driven by the simulation of a social psychological metaphor instead of the survival of the fittest individual. Based on the swarm intelligence theory, this paper discusses the use of PSO with a Quasi-Newton (QN) local search method. The PSO is used to produce good potential solutions, and the QN is used to fine-tune of final solution of PSO. The hybrid methodology is validated for a test system consisting of 13 thermal units whose incremental fuel cost function takes into account the valve-point loading effects.

In this correspondence, an approach based on coevolutionary particle swarm optimization to solve constrained optimization problems formulated as min-max problems is presented. In standard or canonical particle swarm optimization (PSO), a uniform probability distribution is used to generate random numbers for the accelerating coefficients of the local and global terms. We propose a Gaussian probability distribution to generate the accelerating coefficients of PSO. Two populations of PSO using Gaussian distribution are used on the optimization algorithm that is tested on a suite of well-known benchmark constrained optimization problems. Results have been compared with the canonical PSO (constriction factor) and with a coevolutionary genetic algorithm. Simulation results show the suitability of the proposed algorithm in terms of effectiveness and robustness.

This paper presents a hybrid training approach to radial basis function neural networks (RBF-NN). It uses clustering methods to tune the centers of the Gaussian functions used in the hidden layer of a RBF-NN. It also uses particle swarm optimization for centers and spread tuning and the Penrose-Moore pseudo-inverse to adjust the weight's output of the network. Simulations involving this RBF-NN design to identify the chaotic Lorenz' system indicate that the performance of proposed method is better than conventional RBF-NN trained for k-means for multi-step-ahead forecasting.

The classical travelling salesman problem (TSP) is to determine a tour in a weighted graph (that is, a cycle that visits every vertex exactly once) such that the sum of the weights of the edges in this tour is minimal. Hybrid methods, based on nature inspired heuristics, have shown their ability to provide high quality solutions for the TSP. The success of a hybrid algorithm is due to its tradeoff between the exploration and exploitation abilities in search space. This work presents a new hybrid model, based on Particle Swarm Optimization and Fast Local Search, with concepts of Genetic Algorithms, for the blind TSP A detailed description of the model is provided, emphasizing its hybrid features. The control parameters were carefully adjusted and the implemented system was tested with instances from 76 to 2103 cities. For instances up to 439 cities, the best results were less than 1% in excess ofthe known optima. In the average, for all instances, results are 2.538% in excess. Simularion results indicated that the proposed hybrid model performs robustly. These results encourage further research and improvement of the hybrid model to tackle with hard combinatorial problems.

A nonlinear identification approach based on particle swarm optimization (PSO) and Takagi-Sugeno (T-S) fuzzy model for describing dynamical behavior of a thermal-vacuum system is proposed in this paper. Identification of nonlinear systems is an important problem in engineering among what fuzzy models have received particular attention due to their potentialities to approximate nonlinear behavior. Meanwhile PSO is proposed as a method for optimizing the premise part of production rules, least mean squares technique is employed for consequent part of production rules of a T-S fuzzy model. Experimental application using a thermal-vacuum system, used for space environmental emulation and satellite qualification, is analyzed. Numerical results indicate that the PSO succeeded in constructing a T-S fuzzy model for nonlinear identification in this particular application.

Despite the popularity, the tuning aspect of PID controllers is a challenge for researches and plant operators. Various control design methodologies have been proposed in the literature such as auto-tuning, self-tuning and pattern recognition. The main drawback of these methodologies in the industrial environment is the number of tuning parameters to be selected. In this paper, the design of a PID controller, based on the universal model of the plant, is derived. The design characteristic has only one parameter to be tuned. This is a good idea from the viewpoint of plant operators. Fuzzy and neural approaches are used for designing and assessing the efficiency of the PID control in nonlinear plants.

Previous work presented an approach based on coevolutionary particle swarm optimization (Co-PSO) to solve constrained optimization problems formulated as min-max problems. Preliminary results demonstrated that Co-PSO constitutes a promising approach to solve constrained optimization problems. However the difficulty to obtain fine tuning of the solution using a uniform distribution became evident. In this paper, a modified PSO using a Gaussian distribution is applied in the context of Co-PSO. The modified Co-PSO is tested on some benchmark optimization problems and the results show a superior performance compared to the standard Co-PSO.