Hybridization of evolutionary algorithms and local search by means of a clustering method

Abstract

Abstract-This paper presents a hybrid evolutionary algorithm (EA) to solve nonlinear-regression problems. Although EAs have proven their ability to explore large search spaces, they are comparatively inefficient in fine tuning the solution. This drawback is usually avoided by means of local optimization algorithms that are applied to the individuals of the population. The algorithms that use local optimization procedures are usually called hybrid algorithms. On the other hand, it is well known that the clustering process enables the creation of groups (clusters) with mutually close points that hopefully correspond to relevant regions of attraction. Local-search procedures can then be started once in every such region. This paper proposes the combination of an EA, a clustering process, and a local-search procedure to the evolutionary design of product-units neural networks. In the methodology presented, only a few individuals are subject to local optimization. Moreover, the local optimization algorithm is only applied at specific stages of the evolutionary process. Our results show a favorable performance when the regression method proposed is compared to other standard methods.

Noise prediction of a diesel engine fueled with olive pomace oil methyl ester blended with diesel fuel

“…By means of Evolutionary Algorithms (EAs), which are stochastic search algorithms that execute a global search in the input space preventing the fall to local optimum, it is possible to design a near optimal architecture with a great accuracy [30]. This fact and the complexity of the error surface associated with PUNN or RBFNN justify the use of EA to design the structure and adjust the weights of these models.…”

Sensitivity Versus Accuracy in Multiclass Problems Using Memetic Pareto Evolutionary Neural Networks

“…For further details about these mutations and about the generation of the neural networks in the beginning and during the evolutionary process, the reader can consult previous works [47].…”

Combining classification and regression approaches for the quantification of highly overlapping capillary electrophoresis peaks by using evolutionary sigmoidal and product unit neural networks

“…Although the use of PUNNs in this context is very recent, the description of its theoretical basis is outside the scope of this paper, although some current papers from the authors can be considered for further explanations [17,27,28].…”

“…Finally, the stop criterion is reached whenever one of the following two conditions is fulfilled: (i) the algorithm achieves a given number of generations; (ii) there is no improvement for a number of generations either in the average performance of the best 20% of the population or in the fitness of the best individual. For more details about the evolutionary algorithm, see references [27] and [28].…”

“…Although multilayer perceptron (MLP) modeling with sigmoidal unit basis functions (SU) as transfer functions is the most widely used approach [16,[20][21][22][23], radial basis function (RBF) neural networks [19,21] and, more recently, multiplicative neural networks [26], especially the so-called product unit neural networks (PUNNs) [17,27,28] modeling with product unit basis functions (PU) are other interesting choices depending on the analytical problem addressed. In addition to this type of models where all the nodes of the hidden layer have the same type of activation/ transfer functions, hybrid models have also been proposed, where different activation/transfer functions are used for the nodes in the hidden layer.…”