[1/6] Tomorrow (Thursday) at #NeurIPS: Are Greedy Task Orderings Better Than Random in Continual Linear Regression? Q: Do models learn better when consecutive tasks are similar or dissimilar? A: Our analysis suggests that they should be dissimilar! openreview.net/forum?id=8JdP…

[5/6] However, these failure modes disappear when we allow task repetition or use greedy-then-random orderings. In conclusion, we now understand that dissimilarity-guided orderings typically converge faster, and that a bit of randomness can prevent pathological failures.