Deep Generative Dual Memory Network Transforms Continual Learning

A neural architecture combining generative and traditional memory addresses catastrophic forgetting problem in continual learning, where models forget prior tasks when acquiring new ones.