CVPRJun, 2019

元学习中的遗忘学习

TL;DRL2F method proposes task-and-layer-wise attenuation to reduce the influence of prior knowledge for faster adaptation and improved performance in few-shot learning tasks with neural networks.