Supervised Metaplasticity for Big Data: Application to Pollutant Concentrations Forecast

Abstract

Artificial Metaplasticity Multilayer Perceptron is a training algorithm implementation for Artificial Neural Networks inspired in biological metaplasticity property of neurons and Shannon’s information theory. It is based on the hypothesis that a higher amount of information from a Data Set is included in the most atypical data. Using this theory basis a supervised algorithm is developed giving more relevance to the less frequent patterns and subtracting relevance to the more frequent ones. This algorithm has achieved deeper learning on several mutidisciplinar data sets without the need of a Deep Network. The application of this algorithm to a key nowadays environmental problem: the pollutant concentrations prediction in cities, is now considered. The city selected is Salamanca, Mexico, that has been ranked as one of the most polluted cities in the world. The concerning registered pollutants are particles in the order of 10 $\upmu$m or less ($PM_{10}$). The prediction of concentrations of those pollutants can be a powerful tool in order to take preventive measures such as the reduction of emissions and alerting the affected population. In this paper the results obtained are compared with previous recent published algorithms for the prediction of the pollutant concentration. Discussed and conclusions are presented.