Nonparametric randomized response and locally private confidence sets
Project description
NPRR: Nonparametric randomized response
This code implements the nonparametric randomized response (NPRR) mechanism as well as methods for computing locally differentially private confidence intervals and sequences from NPRR's output. The methods are based on the paper "A nonparametric extension of randomized response for private confidence sets" by Ian Waudby-Smith, Zhiwei Steven Wu, and Aaditya Ramdas (2023).
Installation
pip install nprr
Organization
The package is organized into two main submodules:
nprr.mechanisms
implements privacy mechanisms including the NPRR and Laplace mechanisms along with utilities for working with them.nprr.dpcs
implements the confidence intervals and sequences from the paper as well as utilities for working with them.
In addition, the package contains the following submodules:
nprr.cgf
implements some cumulant generating functions (CGF) and CGF-like functions that are used throughout thenprr.dpcs
module.nprr.plotting
implements functions for producing the plots found in the paper.nprr.types
implements some basic type aliases used throughout the package.
Reproducing the plots from the paper
Please begin the following steps from the root of the nprr directory.
- Create and activate a virtual environment using the method of your choice. We used venv, e.g.
python3.9 -m venv venv_dpconc
source venv_dpconc/bin/activate
- Install the package and its dependencies.
python install -e ./
- Enter the plots directory.
cd plots
- Generate figures.
### Figure 1 ###
python hoeffding_eps.py
# -> output file: bounded_beta_1_1_hoeffding_eps.pdf
### Figure 3 ###
python hoeffding_tightness_max.py
# -> output file: bounded_beta_50_50_tightness_max.pdf
### Figure 4 ###
python confint_bounded.py
# -> output file: bounded_beta_50_50_ci.pdf
### Figure 5 ###
python confseq_bounded.py
# -> output file: bounded_beta_50_50_cs.pdf
### Figure 6 ###
python two-sided-running-mean.py
# -> output file: wavy_cs.pdf
### Figure 7 ###
python ab-test.py
# -> output file: wavy_ipw_cs.pdf
Project details
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
File details
Details for the file nprr-0.1.4.tar.gz
.
File metadata
- Download URL: nprr-0.1.4.tar.gz
- Upload date:
- Size: 12.4 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/4.0.2 CPython/3.9.16
File hashes
Algorithm | Hash digest | |
---|---|---|
SHA256 | 60c58e2c50d696f297b44da5f6f40e9f95ff52c990260c54fb03357e084569b2 |
|
MD5 | dc838b45d99bd884bb743634ead445fb |
|
BLAKE2b-256 | 0eebf802a17ff44f7c871a5e5c0ef92c6d8040c194508717c3a3e38868fdad07 |
File details
Details for the file nprr-0.1.4-py3-none-any.whl
.
File metadata
- Download URL: nprr-0.1.4-py3-none-any.whl
- Upload date:
- Size: 12.8 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/4.0.2 CPython/3.9.16
File hashes
Algorithm | Hash digest | |
---|---|---|
SHA256 | d555b626bfb5c6102ff460c95cc98d18912afc147293f0037c2971e1bd4eebfe |
|
MD5 | 3566ce91aa94d82637db51c3543a905d |
|
BLAKE2b-256 | c504e699c551c3702154eb5386633b7a4cfc61cb05885af2571423b9dca4be9e |