Goto

Collaborating Authors

 multi-task baseline


0b5e29aa1acf8bdc5d8935d7036fa4f5-AuthorFeedback.pdf

Neural Information Processing Systems

On the task, all the methods share similar noisy pattern. The43 results show the benefits of adjustingα1 during training. It is shown in [2] that A-GEM has better or comparable44 performance than GEM, so we focus on comparing with A-GEM.