Optimal Private Median Estimation under Minimal Distributional Assumptions

Authors: Christos Tzamos, Emmanouil-Vasileios Vlatakis-Gkaragkounis, Ilias Zadik

NeurIPS 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We present simulation results on synthetic and real data to validate the performance of our proposed Private Median Estimator (PME). We compare our PME with various baselines, including Private Mean Estimator (PME) based on sample mean and the Private Median Estimator (PME) based on sample median.
Researcher Affiliation Academia School of Computer Science, University of Massachusetts Amherst, MA, USA Electrical and Computer Engineering, University of Texas at Austin, TX, USA Department of Computer Science and Engineering, University of California, San Diego, CA, USA
Pseudocode Yes Algorithm 1: Private Median Estimation (Page 4), Algorithm 2: Private Quantile Estimation (Page 5)
Open Source Code No No explicit statement or link providing access to the open-source code for the described methodology was found.
Open Datasets Yes For real-world datasets, we use the Adult dataset [16] and the Bank Marketing dataset [17] from the UCI Machine Learning Repository. [16] Dua, D. and Graff, C. UCI machine learning repository, 2017. [17] S. Moro, P. Cortez, and P. Rita. A data-driven approach to predict the success of bank telemarketing. Decision Support Systems, 62:22–31, June 2014.
Dataset Splits No No specific dataset split information (exact percentages, sample counts, or detailed splitting methodology) for training, validation, or testing was found.
Hardware Specification No No specific hardware details (GPU/CPU models, memory, or detailed computer specifications) used for running experiments were found.
Software Dependencies No No specific ancillary software details (e.g., library or solver names with version numbers) were found.
Experiment Setup Yes For synthetic data experiments, we vary the number of samples n from {1000, 5000, 10000} and the privacy budget ε from {0.1, 0.5, 1.0}. We conduct 100 independent runs for each setting and report the average estimation error.