Multifunctionality in a reservoir computer

Multifunctionality is a well observed phenomenological feature of biological neural networks and considered to be fundamental importance the survival certain species over time. These multifunctional are capable performing more than one task without changing any network connections. In this paper, we investigate how neurological idiosyncrasy can achieved in an artificial setting with modern machine learning paradigm known as “reservoir computing.” A training technique designed enable reservoir computer perform tasks nature. We explore critical effects that changes parameters have on computers’ ability express multifunctionality. also expose existence several “untrained attractors”; attractors dwell within prediction state space were not part training. conduct bifurcation analysis these untrained discuss implications our results.