Created by: vanche

Hi, I wanna ask question to understand your paper. In your paper, you said “when adding a new task we initialize the parameters of the previous one. ” If the previous parameter is initialized, how can the model get knowledge of previous ones? I’m not sure And “Whenever a new task is a introduced, it would be trained with the previous ones.” this means joint learning or something else?

Also I want to know how I can find this procedure in your code