Neural networks for ecology: introduction

A variant of explaining the composition and structure of the neural network is considered, starting from the concept of "regression equation". Focusing on the method more familiar to ecologists (regression analysis), the structural and functional features of the "neuron" and the artificial neural network are shown. The concepts of linear and curvilinear regression, logit, neuron, neural network modeling and algorithms for adjusting the structural and quantitative parameters of models are gradually deepening and expanding. The key terms of the technology under consideration such as covariate, bias, neuron, layer, activation function, training, retraining are defined. In concrete examples, some areas of application of this method in animal ecology are shown. The solution of problems typical for animal ecology of diagnosing the status (sex) of animals by quantitative characteristics and assessing the suitability of certain biotopes for animal habitation with the help of a neural network is considered. A list of references with examples of the use of networks for solving environmental problems is given. Listings of calculations performed in the environment of the R program using functions from the neural net package are given. Data files are attached to the text to perform the training on the presented codes.