Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Con oracle #1391

Merged
merged 25 commits into from
May 9, 2024
Merged

Con oracle #1391

merged 25 commits into from
May 9, 2024

Conversation

AngledLuffa
Copy link
Collaborator

Add an oracle for the in-order-compound transition scheme, along with extensive upgrades to the in-order oracle (although the accuracy gainz are of course not earth-shattering) and some instrumentation such as the ability to add dummy oracle transitions which log the misses in the in-order system.

As incidental changes, include a script to go through the current formatting of conparse results (not expected to change) and some scripts to experiment with different data divisions for building the ensemble. Spoiler: none of the ensemble mechanisms actually helped improve results over just averaging 5 models together

…void making the recall error (the wrong open/open already being a precision error)

Put the ambiguous transition to the bottom of the in-order list.

Add some notes on a slightly improved (?) unambiguous open subtree analysis
…reused in other oracles

Also separate out a find-previous function
Add a partial test as well

Mostly fill out a chart of dynamic oracle transitions for in_order_compound
…taset several times to make ensemble datasets which have a different profile, hopefully resulting in different models
…the transition scheme name a bit less verbose
…of all constituent types (transition types are not covered yet, though, for the case of compound transitions)
Update some stats for when to close an incorrect shift-open
… as to keep statistics on transitions even when they aren't leading to repairs
…ens frequently, it undoes the previous experiments measuring the oracle effectiveness)
Unambiguous solution for some cases as a default repair

Includes two versions of the fix for ambiguous cases
…emed to hurt scores somewhat. Will run another experiment
…closing the extra constituent early or at the end of the surrounding bracket
@AngledLuffa AngledLuffa merged commit 0cd7508 into dev May 9, 2024
1 check passed
@AngledLuffa AngledLuffa deleted the con_oracle branch May 9, 2024 20:54
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

1 participant