We consider different reservoir computers learning and modeling the same chaotic system. We find that by transmitting a common signal, synchronization will be achieved between these superficially distinct reservoir computers. We further show that even when the transmitted signal is contaminated with extensive observational noise, synchronization will still occur. Moreover, utilizing a cascaded synchronization scheme, we can successfully recover a chaotic masking signal hidden in the transmitter. Our findings reveal that synchronization is a common feature of reservoir computers upon learning chaotic systems.
|Journal||Physica A: Statistical Mechanics and its Applications|
|Publication status||E-pub ahead of print - 6 Nov 2019|