Estimation of correlation from limited time-series data using FitzHugh-Nagumo model

Abstract

Since the early 2000s, complex systems have been explored in depth and have been an important topic of research due to tremendous discoveries in real-world networks such as computer, biological, brain, climate, and social networks. Exploring a net work and exploiting its dynamics to generate predictions is sim ple as long as we have information for all of the network’s under lying nodes. But, what if this comprehensive knowledge about the network and its nodes is unavailable, as is the case with real work phenomena? Hereby, employing machine learning tech niques, we offer a collective research of working just with a lim ited number of nodes in a network and using restricted time se ries of these few available nodes to predict correlation matrices. Feed Forward Neural Network is the machine learning algorithm we implemented in this study. Fitzhugh-Nagumo oscillators con trol the network’s dynamics, and the coupled dynamics of these oscillators are utilised to generate synthetic data i.e. time series of the underlying nodes of the network.