Repository navigation
Add NestedTabularSynthesizer for typed/nested tabular data under dpsynth/google/research/activity_logs. - #83
Merged
Merged
Conversation
copybara-service
Bot
force-pushed
the
cl/940786768
branch
from
July 8, 2026 02:03
f838c06 to
a91a9e4
Compare
copybara-service
Bot
changed the title
copybara-service
Bot
force-pushed
the
cl/940786768
branch
7 times, most recently
from
July 13, 2026 10:46
796235b to
4339268
Compare
copybara-service
Bot
force-pushed
the
cl/940786768
branch
2 times, most recently
from
July 27, 2026 08:16
864462c to
8788708
Compare
copybara-service
Bot
force-pushed
the
cl/940786768
branch
from
August 3, 2026 14:50
8788708 to
681bbc1
Compare
copybara-service
Bot
changed the title
copybara-service
Bot
force-pushed
the
cl/940786768
branch
from
August 12, 2026 16:09
681bbc1 to
21cc4bb
Compare
copybara-service
Bot
changed the title
copybara-service
Bot
force-pushed
the
cl/940786768
branch
from
August 13, 2026 16:41
21cc4bb to
5d08ec9
Compare
…nth/google/research/activity_logs. Introduces NestedTabularSynthesizer in dpsynth/google/research/activity_logs that handles datasets where each record has a type and different types have different attribute schemas. Two-model architecture: - Model 1 (shared): learns joint distribution over record types + shared attributes using a single TabularSynthesizer. - Model 2 (per-type): learns type-specific attribute distributions via independent TabularSynthesizers under parallel composition. Privacy accounting: sequential composition of Model 1 + Model 2 (parallel composition for Model 2). The detail level reports a ZCDpEvent at the calibrated rho rather than using an arbitrary per-type dp_event, since different schemas may produce different dp_events that could undercount under PLD accounting. Input/output format: dict[str, pd.DataFrame] keyed by type name. PiperOrigin-RevId: 964138819
copybara-service
Bot
force-pushed
the
cl/940786768
branch
from
August 13, 2026 16:48
5d08ec9 to
280033e
Compare
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
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.
Add NestedTabularSynthesizer for typed/nested tabular data under dpsynth/google/research/activity_logs.
Introduces NestedTabularSynthesizer in dpsynth/google/research/activity_logs
that handles datasets where each record has a type and different types have
different attribute schemas.
Two-model architecture:
attributes using a single TabularSynthesizer.
independent TabularSynthesizers under parallel composition.
Privacy accounting: sequential composition of Model 1 + Model 2 (parallel
composition for Model 2). The detail level reports a ZCDpEvent at the
calibrated rho rather than using an arbitrary per-type dp_event, since
different schemas may produce different dp_events that could undercount
under PLD accounting.
Input/output format: dict[str, pd.DataFrame] keyed by type name.