
Attention-Enhanced Reservoir Computing as a Multiple Dynamical System Approximator
Reservoir computing has proven effective for tasks such as time-series prediction, particularly in the context of chaotic systems. However, conventional reservoir computing frameworks often face challenges in achieving high prediction accuracy and adapting to diverse dynamical problems due to their reliance on fixed weight structures. A concept of an attention-enhanced reservoir computer has been proposed, which integrates an attention mechanism into the output layer of the reservoir computing …