Open 610v4nn1 opened 1 year ago
The list of supervised strategies may be the closest thing to what you are asking. Keep in mind that it's not a one-to-one mapping with the literature. For example, Replay
is a generic strategy which many policy, so it correspond to multiple rehearsal strategies.
We should probably add a list of strategies to the website.
Thanks for the pointer. I think that it will be hard to document all the possible combinations but just having a few example with 1:1 mapping to very competitive algorithms would already help.
Currently, three implemented strategies are not on the list:
LaMAML -> meta-learning-based MER -> meta-learning-based ER-ACE -> online and batch version
They are all under avalanche/training/supervised/
Thank you both for the hints
I am looking for a list of supported CL algorithms but I was not able to find it in the documentation. In the avalanche-baselines project I saw some well-known baselines but they don't seem to cover everything available in the library (?).
I am just looking for strong baselines to be used in an empirical comparison and I think a table/list of available algorithms will help others as well.