fix(hpo): exclude DOMIAS from the optimization objective - #13
Merged
Conversation
DOMIAS estimates membership-inference risk with a Gaussian KDE over SynthCity's encoded data. On wide datasets such as LORIS, the encoded feature space can exceed the number of synthetic rows or produce a rank-deficient covariance matrix, causing SciPy to fail with a singular covariance error. SynthCity can catch that metric failure and omit the corresponding result row. Keeping DOMIAS in HPO therefore leaves trials with incomplete and potentially incomparable privacy scores, which can influence model selection without a valid DOMIAS measurement. Retain identifiability_score as the HPO privacy objective and remove DOMIAS from the default and shipped HPO profiles. Leave the final-evaluation catalog unchanged so this interim fix does not redefine the existing evaluation protocol. Add regression tests covering both HPO exclusion and final-evaluation retention. Because the HPO objective changed, existing Optuna studies and cached best parameters must not be reused for new comparisons. Bump the package version to 0.7.4. Fixes #3
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
DOMIAS estimates membership-inference risk with a Gaussian KDE over SynthCity's encoded data. On wide datasets such as LORIS, the encoded feature space can exceed the number of synthetic rows or produce a rank-deficient covariance matrix, causing SciPy to fail with a singular covariance error.
SynthCity can catch that metric failure and omit the corresponding result row. Keeping DOMIAS in HPO therefore leaves trials with incomplete and potentially incomparable privacy scores, which can influence model selection without a valid DOMIAS measurement.
Retain identifiability_score as the HPO privacy objective and remove DOMIAS from the default and shipped HPO profiles. Leave the final-evaluation catalog unchanged so this interim fix does not redefine the existing evaluation protocol. Add regression tests covering both HPO exclusion and final-evaluation retention.
Because the HPO objective changed, existing Optuna studies and cached best parameters must not be reused for new comparisons. Bump the package version to 0.7.4.
Fixes #3