We have several learners, but in our current plan only subset can currently make use of our improved continuous feature matching. Extending this to other learners is out of scope for the foreseeable future. However, it seems worth documenting this as "not implemented" since the lack of an implementation would otherwise be hard to notice without running experiments.
I think there are two related things we would need to handle for this issue if we needed to address it:
Figure out and implement a way of doing updates. This means figuring out when/how to call confirm_match() on some pattern match info, or how we would update continuous nodes if we didn't do it by confirming matches the way we do now. There may be obvious times and places to do this for some learners; for others it may be less obvious.
Handle the new continuous predicate node type properly for pattern-pattern matching and for the pattern generalization some non-subset learners do.
We have several learners, but in our current plan only subset can currently make use of our improved continuous feature matching. Extending this to other learners is out of scope for the foreseeable future. However, it seems worth documenting this as "not implemented" since the lack of an implementation would otherwise be hard to notice without running experiments.
I think there are two related things we would need to handle for this issue if we needed to address it:
confirm_match()
on some pattern match info, or how we would update continuous nodes if we didn't do it by confirming matches the way we do now. There may be obvious times and places to do this for some learners; for others it may be less obvious.