Finding reducts is one of the key problems in the increasing applications of rough set theory, which is also one of the bottlenecks of the rough set methodology. The population-based reduction approaches are attractive to find multiple reducts in the decision systems, which could be applied to generate multi-knowledge and to improve decision accuracy. In this paper, we design a multi-swarm synergetic optimization algorithm (MSSO) for rough set reduction and multi-knowledge extraction. It is a multi-swarm based search approach, in which different individual trends to be encoded to different reduct. The approach discovers the best feature combinations in an efficient way to observe the change of positive region as the particles proceed throughout the search space. The performance of our approach is evaluated and compared with Standard Particle Swarm Optimization (SPSO) and Genetic Algorithms (GA). Empirical results illustrate that the approach can be applied for multiple reduct problems and multi-knowledge extraction effectively.
The multilayer perceptron model has been suggested as an alternative to conventional approaches, and can accurately forecast time series. Additionally, several novel artificial neural network models have been proposed as alternatives to the multilayer perceptron model, which have used (for example) the generalized- mean, geometric mean, and multiplicative neuron models. Although all of these artificial neural network models can produce successful forecasts, their aggregation functions mean that they are negatively affected by outliers. In this study, we propose a new multilayer, feed forward neural network model, which is a robust model that uses the trimmed mean neuron model. Its aggregation function does not depend on outliers. We trained this multilayer, feed forward neural network using modied particle swarm optimization. We applied the proposed method to three well-known time series, and our results suggest that it produces superior forecasts when compared with similar methods.
This paper presents a hybrid probabilistic neural network (PNN) and particle swarm optimization (PSO) techniques to predict the soil liquefaction. The PSO algorithm is employed in selecting the optimal smoothing parameter of the PNN to improve the forecasting accuracy. Seven parameters such as earthquake magnitude, normalized peak horizontal acceleration at ground surface, standard penetration number, penetration resistance, relative compaction, mean grain diameter and groundwater table are selected as the evaluating indices. The predictions from the PSO-PNN model were compared with those from two models: backpropagation neural network (BPNN) model and support vector machine (SVM) model. The study concluded that the proposed PSO-PNN model can be used as a reliable approach for predicting soil liquefaction.
Purpose of this work is to show that the Particle Swarm Optimization Algorithm may improve the results of some well known Machine Learning methods in the resolution of discrete classification problems. A binary version of the PSO algorithm is used to obtain a set of logic rules that map binary masks (that represent the attribute values), to the available classes. This algorithm has been tested both in a single pass mode and in an iterated mode on a well-known set of problems, called the MONKS set, to compare the PSO results against the results reported for that domain by the application of some common Machine Learning algorithms.
An artificial neural network (ANN) based on particle swarm optimization (PSO) was developed for the time series prediction. This hybrid ANN+PSO algorithm was applied on Mackey-Glass series in the short-term prediction x(t+6) and the long-term prediction x(t+84), from the current value x(t) and the past values: x(t-6), x(t-12), x(t-18). Four cases were studied, alternating the timedelay parameter as 17 or 30. Also, the first four largest Lyapunov exponents were obtained for different time-delay. Simulation shows that this ANN+PSO method is a very powerful tool for making prediction of chaotic time series.
In this article, a technique called Meta-Optimization is used to enhance the effectiveness of bio-inspired algorithms that solve antenna array synthesis problems. This technique consists on a second optimization layer that finds the best behavioral parameters for a given algorithm, which allows to achieve better results. Bio-inspired computational methods are useful to solve complex multidimensional problems such as the design of antenna arrays. However, their performance depends heavily on the initial parameters. In this paper, the distances between antenna array elements are calculated in order to reduce electromagnetic interference from undesired sources. The results are compared to previous works, showing an improvement on the performance of bio-inspired optimization algorithms such as Particle Swarm Optimization and Differential Evolution. These results are found to be statistically significant based on the Wilcoxon's rank sum test as compared to these methods using the standard parameters proposed in the literature. Furthermore, graphical representations of the Meta-Optimization process called meta-landscapes are presented, showing the behavior of these algorithms for a range of different parameters, providing the best parameter combinations for each antenna problem.
Diabetes mellitus (DM) is a disease affecting millions of people worldwide, and its medical care brings an economic wear to patients and public health systems. Many efforts have been made to deal with DM, one of them is the full-automation of insulin delivery. This idea consists in design a closed-loop control system to maintain blood glucose levels (BGL) within normal ranges. Dynamic models of glucose-insulin-carbohydrates play an important role in synthesis of control algorithms, but also in other aspects of DM care, such as testing glucose sensors, or as support systems for health care decisions. Therefore, there are several mathematical models reproducing glycemic dynamics of DM, most of them validated with nominal parameters of standardized patients. Nevertheless, individual patient-oriented models could open the possibility of having closed-loop personalized therapies. This problem can be addressed through the information provided by open-loop therapy based on continuous glucose monitoring and subcutaneous insulin infusion. This paper considers the problem of identifying particular parameters of a compartmental model of glucose-insulin dynamics in DM; the goal is fitting the model response to historical data of a diabetic patient collected during a time period of her/his daily life. At this time, Sorensen model is one of the most complete compartmental models representing the complex dynamics of the glucose-insulin metabolism. This is a system of 19 ordinary differential equations (ODEs), thus the identification of its parameters is a non-easy task. In this contribution, parameter identification was performed via three evolutionary algorithms: differential evolution, ant colony optimization and particle swarm optimization. The obtained results show that evolutionary algorithms are powerful tools to solve problems of parametric identification. Also, a comparative analysis of the three algorithms was realized throw a wilcoxon sign-rank test, in which colony optimization had the better performance. The model obtained with the estimated parameters could be used to in type 1 diabetes mellitus (T1DM) care, such as in the design of full-automation of insulin infusion.