Trade-off between memory capacity and nonlinearity

Ascertain whether memory capacity and nonlinearity in reservoir computing present fundamental trade-offs and characterize their quantitative relationship to guide substrate design and task performance.

Background

Reservoir performance depends on both nonlinearity and memory properties. Prior work suggests possible trade-offs, but a general characterization remains unsettled. Clarifying whether increasing nonlinearity erodes memory capacity (or vice versa) would inform how to tune physical and computational reservoirs for diverse prediction tasks.