Regression with Label Differential Privacy

Authors: Badih Ghazi, Pritish Kamath, Ravi Kumar, Ethan Leeman, Pasin Manurangsi, Avinash Varadarajan, Chiyuan Zhang

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

Reproducibility Variable Result LLM Response
Research Type Experimental We carry out a thorough experimental evaluation on several datasets demonstrating the efficacy of our algorithm.
Researcher Affiliation Industry Email: {badih.ghazi, ravi.k53}@gmail.com, {pritishk, ethanleeman, pasin, avaradar, chiyuan}@google.com
Pseudocode Yes Algorithm 1 RR-on-BinsΦ ε. [...] Algorithm 2 Compute optimal Φ for RR-on-BinsΦ ε. [...] Algorithm 3 Labels Party s Randomizer Label Randomizerε1,ε2.
Open Source Code No The paper does not provide concrete access to source code for the methodology described in this paper.
Open Datasets Yes The Criteo Sponsored Search Conversion Log Dataset (Tallis & Yadav, 2018) is publicly available from https://ailab. criteo.com/criteo-sponsored-search-conversion-log-dataset/. [...] The 1940 US Census dataset has been made publicly available for research since 2012
Dataset Splits No The paper mentions "80% 20% train test splits" but does not specify a validation dataset split.
Hardware Specification No The paper does not provide specific hardware details (e.g., GPU/CPU models, processor types) used for running its experiments.
Software Dependencies No The paper mentions software like "Tensor Flow Privacy" and "Py Torch Opacus" but does not specify version numbers for these or other relevant libraries/solvers.
Experiment Setup Yes We use learning rate 0.001 with cosine decay (Loshchilov & Hutter, 2017), batch size 8192, and train for 50 epochs. [...] We use learning rate 0.001 with cosine decay (Loshchilov & Hutter, 2017), batch size 8192, and train for 200 epochs.