Techniques of Machine Learning Mixed with Metaheuristic Methods

Summary

This chapter covers the hybridization of machine learning techniques and metaheuristic methods. Artificial Neural Networks (ANNs) are probably the most popular machine learning technique nowadays. As the ANNs find their way into different applications, several architectures and training techniques have been proposed. It is safe to say that the efficacy of an ANN relies on choosing an appropriate architecture for the problem at hand and finding the combination of weights that supplies the network with sufficient generalization capability. Choosing the correct network architecture and training process has proven difficult to automate, remaining primarily a craft where the designer's expertise plays a central role. To illustrate the hybridization of machine learning techniques and metaheuristic methods, this chapter presents the training of a Feedforward Neural Network (FNN) through the Flower Pollination Algorithm (FPA). The chapter aims to transmit to the reader, through a basic example, the high potential of techniques hybridization.