I was studying meta-learning recently, and there was an algorithm that used "fast-weights" to facilitate the training speed. Can someone explain to me how they act
