Implicit biases in multitask and continual learning from a backward error analysis perspective

Open in new window