Artificial Neural Networks and Efficient Optimization Techniques for Applications in Engineering

This chapter proposal describes some artificial neural network (ANN) neuromodeling techniques used in association with powerful optimization tools, such as natural optimization algorithms and wavelet transforms, which can be used in a variety of applications in Engineering, for example, Electromagnetism (Cruz, 2009), Signal Processing (Peixoto et al., 2009b) and Pattern Recognition and Classification (Magalhães et al., 2008). The application of ANN models associated with RF/microwave devices (Cruz et al., 2009a, 2009b; Silva et al., 2010a) and/or pattern recognition (Lopes et al., 2009) becomes usual. In this chapter, we present neuromodeling techniques based on one or two hidden layer feedforward neural network configurations and modular neural networks − trained with efficient algorithms, such as Resilient Backpropagation (RPROP) (Riedmiller & Braun, 1993), Levenberg-Marquardt (Hagan & Menhaj, 1999) and other hybrid learning algorithms (Magalhães et al., 2008), in order to find the best training algorithm for such investigation, in terms of convergence and computational cost. The mathematical formulation and implementation details of neural network models, wavelet transforms and natural optimization algorithms are also presented. Natural optimization algorithms, which are stochastic population-based global search methods inspired in nature, such as genetic algorithm (GA) and particle swarm optimization (PSO) are effective for optimization problems with a large number of design variables and inexpensive cost function evaluation (Kennedy & Eberhart, 1995; R. Haupt & S. Haupt, 2004). However, the main computational drawback for optimization of nonlinear devices relies on the repetitive evaluation of numerically expensive cost functions (Haupt & Werner, 2007; Rahmat-Samii, 2003). Finding a way to shorten the optimization cycle is highly desirable. In case of GA, for example, several schemes are available in order to improve its performance, such as: the use of fast full-wave methods, micro-genetic algorithm, which aims to reduce the population size, and parallel GA using parallel computation (R. Haupt & S. Haupt, 2004; Haupt & Werner, 2007). Therefore, this chapter