tensorflow / tensorflow/privacy
compute_noise_from_budget for PATE
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 2k
- Forks
- 477
- Avg merge
- 22h 12m
- Merged PRs (30d)
- 1
Description
As mentioned in https://github.com/tensorflow/privacy/pull/143#issue-524365572, the noise multiplier is an important "hyperparameter" to choose as it would affect the outcome of the target privacy budget. However due to PATE being dependent on the teacher's output, is there a method to calculate the noise multiplier for it? or is it totally impossible?
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start by reading the discussion in pull request 143 and locating the PATE implementation in this repository. Determine whether the teacher output provides enough information to derive a noise multiplier from a target privacy budget, then define and validate the expected behavior before proposing an implementation.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- machine-learning, security
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 20/100