Search

Real-world parameter identification problems require determining the bounds that cover the unknown solutions. This paper presents an adaptive differential evolution algorithm with a bound adjustment strategy (ADEBAS) for solving nonlinear parameter identification problems. The adjustment strategy detects the parameter-bound violations of mutant vectors during the evolution process and gradually extends the bounds. The algorithm adaptively uses two mutation strategies and two ranges of crossover rate to balance the population diversity and convergence speed. Experimental results show that ADEBAS can solve 24 nonlinear regression tasks from the National Institute of Standards and Technology benchmark with accurate estimation and reliability. It also outperforms the compared methods on real-world parameter identification problems.

Artificial neural networks are essential intelligent tools for various learning tasks. Training them is challenging due to the nature of the data set, many training weights, and their dependency, which gives rise to a complicated high-dimensional error function for minimization. Thus, global optimization methods have become an alternative approach. Many variants of differential evolution (DE) have been applied as training methods to approximate the weights of a neural network. However, empirical studies show that they suffer from generally fixed weight bounds. In this research, we propose an enhanced differential evolution algorithm with adaptive weight bound adjustment (DEAW) for the efficient training of neural networks. The DEAW algorithm uses small initial weight bounds and adaptive adjustment in the mutation process. It gradually extends the bounds when a component of a mutant vector reaches its limits. We also experiment with using several scales of an activation function with the DEAW algorithm. Then, we apply the proposed method with its suitable setting to solve function approximation problems. DEAW can achieve satisfactory results compared to exact solutions.

In this research, an improved differential evolution algorithm with a restart technique (DE-R) is designed for solutions of systems of nonlinear equations which often occurs in solving complex computational problems involving variables of nonlinear models. DE-R adds a new strategy for mutation operation and a restart technique to prevent premature convergence and stagnation during the evolutionary search to the basic DE algorithm. The proposed method is evaluated on various real world and synthetic problems and compared with the recently developed methods in the literature. Experiment results show that DE-R can successfully solve all the test problems with fast convergence speed and give high quality solutions. It also outperforms the compared methods.

Clustering is an essential data exploration technique applied to many disciplines and applications such as data mining, image processing, bioinformatics, and machine learning. A clustering method identifies hidden patterns in a dataset and combines similar data points into clusters. The problems are challenging when they have many data points, attributes, and cluster partitions. In this paper, we propose an adaptive differential evolution with an archive strategy (ADEAS) to find candidate centroids and minimize their intra-cluster distance for solving partitional clustering problems. The archiving strategy stores inferior solutions during the selection operation to increase population diversity and create directions for guiding the search. We validate the proposed algorithm with several well-known methods using the UCI datasets. The results show that ADEAS outperforms the compared methods. [ABSTRACT FROM AUTHOR]