We present a framework for physical local learning in metamaterials based on a linear flow network with symmetric and antisymmetric components. This model extends previous work on steady-state ...