Private Statistical Estimation of Many Quantiles

Authors: Clément Lalanne, Aurélien Garivier, Rémi Gribonval

ICML 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Figure 1. Numerical performance of the different private estimators
Researcher Affiliation Academia 1Univ Lyon, Ens L, UCBL, CNRS, Inria, LIP, F-69342, LYON Cedex 07, France 2Univ. Lyon, ENS de Lyon, UMPA UMR 5669, 46 all ee d Italie, F-69364 Lyon cedex 07.
Pseudocode No The paper describes algorithms such as QExp, Ind Exp, and Rec Exp in prose, often referencing original papers for full details, but does not provide structured pseudocode or algorithm blocks within its content.
Open Source Code No The paper mentions that QExp 'is implemented in many DP software libraries (Allen; IBM)' but does not provide any statement or link for code released by the authors of this paper for their described methodology.
Open Datasets No For the experiments, we benchmarked the different estimators on beta distributions, as they allow to easily tune the Lipschitz constants of the densities, which is important for characterizing the utility of the histogram estimator.
Dataset Splits No The paper does not provide specific dataset split information (exact percentages, sample counts, citations to predefined splits, or detailed splitting methodology) for its experiments.
Hardware Specification No The paper does not provide specific hardware details (exact GPU/CPU models, processor types with speeds, memory amounts, or detailed computer specifications) used for running its experiments.
Software Dependencies No The paper does not provide specific ancillary software details, such as library or solver names with version numbers, needed to replicate the experiment.
Experiment Setup Yes We estimate the quantiles of orders p = 1 4 + 1 2(m+1), . . . , 1 4 + m 2(m+1) for different values of m, n = 10000, ϵ = 0.1, q is the private estimator, and E is estimated by Monte-Carlo averaging over 50 runs. The histogram is computed on 200 bins.