Inspired by human brain for multi-task continual learning, a generalized photonic neuromorphic architecture (L2ONN) is proposed to model physical-driven light sparsity and parallelism, towards reconfigurable and scalable lifelong learning.
