Machine learning applications in Chaotic Rossler oscillators

Abstract

In real-world dynamical systems, partial information about the state of a system is often only available, and predicting the future time evolution of the system from partial time series leads to significant errors in the long term. A correlation matrix of time series is a valuable measure in understanding a system’s state. Learning the system’s state has many applications in brain research and Earthquake prediction to set up warning systems. Machine learning techniques are popular statistical tools used for prediction in all analytical fields. Using these techniques to predict a system’s state from limited time-series information of a few nodes is a novel way to study the system. To this extent, we have devised a model to predict the correlation matrices from limited time series information of a few nodes.