Special issue on information processing and machine learning for applications of engineering

Abstract

This special issue aims to cover some problems related to information processing and machine learning for applications of Engineering. The origin of this volume is in the IEEE 17th International Conference on Intelligent Engineering Systems, INES׳13, held at Heredia (Costa Rica) on June 19–21, 2013. The INES Series started in 1997 have become an annual event whose aim is to provide researchers and practitioners from industry and academia with a platform to report on recent developments in the area of computational intelligence. INES 2013 focuses on the application of state-of-the-art intelligent techniques to engineering systems. A selected choice of papers based on the revision processes and presentations delivered at INES׳13 has given rise to this issue of Neurocomputing.

The topics of the special issue are an interesting and active field and 9 papers have been submitted, selected from up to 70 papers. After at least two rounds of reviews, 5 papers were selected for publication. The papers can be categorized into two clusters based on machine learning and information processing for engineering. We summarize the papers as follows: (1) Machine learning: Reiner and Wilamowski showed an artificial neural network as a very efficient universal approximation, in particular, the construction of a Single Layer Feedforward Networks (SLFN) architecture using radial basis function (RBF) neurons. There are many algorithms that are used to construct or train networks to solve function approximation problems. In this work, an algorithm which is a modification of the Incremental Extreme Learning Machine (I-ELM) family of algorithms is proposed. The proposed algorithm eliminates randomness in the learning process with respect to center positions and widths of the RBF neurons. To do this, the input with the highest error magnitude is saved during error calculation and then used as the center for the next incrementally added neuron. Then the radius of the new neuron is iteratively chosen using Nelder–Mead׳s Simplex method. This allows the universal approximation properties of I-ELM to be preserved while greatly reducing the sizes of the trained RBF networks. Ben Arab Taher et al. proposed an approach for examining the problem of estimating the parameters of the proposed Diagonal of the Modified Riesz Distribution (DMRD). This work estimates the parameters of the DMRD through a Bayesian approach associated with Monte Carlo methods. The proposed approach is compared via a simulation study with maximum likelihood method and method of moments by calculating the Mean Square Error (MSE).

(2) Information processing: Alonso et al. aimed an approach to analyze the improvements are observed in the glottal excitation synthesizers when the possible manifestations of non-linear behavior are characterized in glottal excitation. This work proposes a new model based on the modification of a classic glottal excitation synthesizer and to study the improvements regarding different glottal excitation synthesizers. The proposed model tries to improve the naturalness of the synthesized voice using the synthesis of the sub-harmonics. The proposed model is included in a generic synthesizer of sustained vowels in order to get an assessment about the quality of the synthesis of different qualities of voice, where speakers with pathologies in the phonatory system are used to simulate the behavior of low quality voices. The different models are adjusted using genetic algorithms. The assessment of the different glottal excitation synthesizers is obtained using an objective measure of similarity between the original signals and synthesized signals based on temporal and spectral measurements. In addition, the quality of the proposed glottal excitation model is evaluated with a study of subjective perception. Pozo-Baños et al. investigated a pollen grain identification approach. A combination of geometrical and texture characteristics are proposed as pollen grain discriminative features as well as the usage of the most popular feature extraction techniques. Multi-Layer Neural Network and Least Square Support Vector Machine (LS-SVM) with Radial Basis Function were used as classifier systems. K-Fold and Hold-Out cross-validation techniques were applied in order to achieve reliable results. When testing with a 17-species database, the combination of the proposed set of features processed by Linear Discriminant Analysis and the LS-SVM has provided the best performance. López-de-Ipiña et al. presented an automatic Analysis of Emotional Response (AAER) in spontaneous speech based on classical and new emotional speech features: Emotional Temperature (ET) and fractal dimension (FD). This is a pre-clinical study aiming to validate tests and biomarkers for future diagnostic use. The method has the great advantage of being noninvasive, low cost, and without any side effects. The AAER shows very promising results for the definition of features useful in the early diagnosis of AD.