Artificial Neural Networks: Multilayer Perceptron for Ecological Modeling

Artificial neural networks (ANNs) are biologically inspired computational networks. Among the various types of ANNs, in this chapter, we focus on multilayer perceptrons (MLPs) with backpropagation learning algorithms. MLPs, the ANNs most commonly used for a wide variety of problems, are based on a supervised procedure and comprise three layers: input, hidden, and output. We discuss various aspects of MLPs, including structure, algorithm, data preprocessing, overfitting, and sensitivity analysis. In addition, we outline the advantages and disadvantages of MLPs and recommend their usage in ecological modeling. Finally, an example demonstrating the practical application of MLP in ecological models is presented.