Synchronizing hyperchaotic subsystems with a single variable: A reservoir computing approach

Abstract We adopt the machine learning technique known as “reservoir computing” to investigate hyperchaos synchronism. Interestingly, we find that by virtue of this approach, synchronization of hyperchaotic subsystems is available via only one dynamical variable. Specifically, we show that by sending just a single variable, synchronization of hyperchaotic subsystem and its learned reservoir computer can be achieved. In the same fashion, one can further observe synchronism in the trained reservoir computers even when their learned hyperchaotic subsystems are mismatched. Moreover, we demonstrate that the synchronization of interest is robust in the presence of relatively high levels of extensive noise. Our work provides an alternative way of realizing synchronization in hyperchaotic systems.