Nonparametric Extensions of Randomized Response for Private Confidence Sets
Authors: Ian Waudby-Smith, Steven Wu, Aaditya Ramdas
ICML 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | For example, our results yield private analogues of Hoeffding s inequality in both fixed-time and time-uniform regimes. We extend these Hoeffding-type CSs to capture time-varying (non-stationary) means, and conclude by illustrating how these methods can be used to conduct private online A/B tests. The efficiency gains that result from this approach are illustrated in Figures 4 and 5. |
| Researcher Affiliation | Academia | Ian Waudby-Smith 1 Zhiwei Steven Wu 1 Aaditya Ramdas 1 1Carnegie Mellon University. Correspondence to: Ian Waudby Smith <ianws@cmu.edu>. |
| Pseudocode | Yes | Algorithm 1 Sequentially interactive Laplace mechanism. Algorithm 2 Nonparametric randomized response (NPRR) |
| Open Source Code | Yes | A Python package implementing our methods as well as code to reproduce the figures can be found on Git Hub at github.com/Wannabe Smith/nprr. |
| Open Datasets | No | The paper mentions evaluating on data from a "uniform distribution" and "Beta(50, 50) distribution" in figures, which are standard statistical distributions, but does not provide specific links, citations, or instructions to access a dataset based on these. It discusses bounded random variables in general. It does not provide concrete access information to a publicly available dataset. |
| Dataset Splits | No | The paper does not explicitly mention training/validation/test splits, instead focusing on theoretical bounds and empirical performance on distributions. |
| Hardware Specification | No | No specific hardware details (GPU/CPU models, memory amounts) are mentioned in the paper for running experiments. |
| Software Dependencies | No | The paper mentions 'A Python package implementing our methods' but does not specify Python version or any other library versions. Therefore, it does not provide specific ancillary software details with version numbers. |
| Experiment Setup | No | The paper discusses various parameters for its proposed methods (e.g., ε, r, G, λt, β, D, c) and suggests how to choose them (e.g., 'we set c 0.1', 'we set D 30'), which are akin to hyperparameters. However, it does not provide a dedicated 'Experimental Setup' section with a comprehensive list of all parameters used for figures or specific training configurations in one place. |