Meteorological patterns recognition using Artificial Neural Networks programmed with the Swish activation function

Abstract

Artificial neural networks are a set of tools that are widely used for the information classification. Its expansion within artificial intelligence has been due to its use in the Machine Learning area. A fundamental part of the artificial neural networks algorithm is the so-called activation function, the above because it is the part that triggers the process as a whole and due to its result the neuron/perceptron sends its outputs. Back-propagation activation function of an artificial neural network is also described; this is artificial neural network with a simpler functioning whose adaptation has made it especially attractive to pattern recognition; also, a different algorithm such as Swish is introduced. As part of the pattern recognition study, three wind classifications present on the Mexican Republic Atlantic coast are formed, each group is made up of graphic files referring to meteorological maps with wind indicators in order to feed the network and as new maps are generated, the Artificial Neural Network will be an aid in the meteorological patterns detection.