Data-Specific Hyper-Parameter Design: A Paradigm Shift in Reservoir Computing

Reservoir computing performance improves by replacing random reservoir selection with geometry-driven design that aligns reservoir state increments within a cone around input-determined vector subspaces, eliminating trial-and-error hyperparameter tuning for inputs from deterministic dynamical systems.