Convolution-based machine learning to attenuate Covid-19's infections in large cities

Abstract

ABSTRACT In this paper a nonlinear mathematical model based at convolution theory and translated in terms of Machine Learning philosophy is presented. In essence, peaks functions are assumed as the pattern of rate of infections at large cities. In this manner, once the free parameters of theses patterns are identified then one proceeds to engage to the well-known Mitchell's criteria in order to construct the algorithm that would yield the best estimates as to carry out social intervention as well as to predict dates about the main characteristics of infection's distributions. The distributions are modeled by the Dirac-Delta function whose spike property is used to make the numerical convolutions. In this manner the parameters of Dirac-Delta function's argument are interpreted as the model parameters that determine the dates of social regulation such as quarantine as well as the possible date of end of first wave and potential periods of the beginning of a second one. The theoretical and computational approach is illustrated with a case of outbreak depending on free parameters simulating the implementation of new rules to detain the infections.