Generate Binomial Noise Function #1082
Conversation
ipa-core/src/protocol/ipa_prf/prf_sharding/feature_label_dot_product.rs
Outdated
Show resolved
Hide resolved
|
@martinthomson I think I addressed most of your earlier feedback. But two things I'd like to put in separate PRs
Working to resolve a few more clippy errors and then hoping we can merge this PR. |
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #1082 +/- ##
==========================================
+ Coverage 90.95% 91.00% +0.05%
==========================================
Files 184 185 +1
Lines 26079 26216 +137
==========================================
+ Hits 23720 23858 +138
+ Misses 2359 2358 -1 ☔ View full report in Codecov by Sentry. |
|
All tests are passing now |
@martinthomson I looked into this some more and somewhat surprisingly it doesn't seem that (at least in the 1 dimensional case) we can decrease the error by decreasing p with the same epsilon. See |
| #[cfg(test)] | ||
| pub async fn gen_binomial_noise<'ctx, const B: usize, OV>( | ||
| ctx: UpgradedSemiHonestContext<'ctx, NotSharded, Boolean>, | ||
| _never_used_input: BitDecomposed<Replicated<Boolean, B>>, |
There was a problem hiding this comment.
@bmcase - can you explain what's going on with _never_used_input?
This PR includes the generation of Binomial noise in MPC. There will be a separate PR to apply the noise to outputs.
More tests will be added to this PR to test the
gen_binomial_noisefunction but this is anyway a working version of it if folks want to start reviewing.This PR will also get smaller after #1076, #1067 and #1072 are merged separately and rebased on.