MPI Campus Seminar: Network hierarchies to store memories

Most successful networks, including economic, societal, transport and vasculature networks, are dynamic.Continuous adaptation of network links ensures optimal network performance when challenged with dynamic external influences. Yet, how can continuous adaptation in a dynamic environment be aligned with the observation that stimuli trigger permanent changes in adaptive networks? We show that adaptation dynamics allow a network to memorize information about the position of an applied stimulus within its network morphology. Both our numerical simulations and analytical calculations identify that the irreversible dynamics of vanishing network links encode memory. Our analytical theory faithfully predicts the role of all system parameters during memory formation, particularly those parameter values which prevent memory formation. This not only opens an entirely new perspective on adaptive networks such as our vasculature but provides an analytically tractable theory of memory formation in disordered systems.