Differentially Private Decomposable Submodular Maximization

Authors: Anamay Chaturvedi, Huy Lê Nguyễn, Lydia Zakynthinou6984-6992

AAAI 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We complement our theoretical bounds with experiments demonstrating improved empirical performance.
Researcher Affiliation Academia Khoury College of Computer Sciences, Northeastern University, Boston, Massachusetts 02115
Pseudocode Yes Algorithm 1 Private Continuous Greedy
Open Source Code Yes The code and dataset used for our experiments are available at https://github.com/Anamay-Chaturvedi/Differentially-privatedecomposable-submodular-optimization
Open Datasets Yes We use the same dataset of coordinates of Uber pick-ups2. 2https://www.kaggle.com/fivethirtyeight/uber-pickups-in-newyork-city.
Dataset Splits No The paper does not provide specific train/validation/test dataset splits. It describes a resampling strategy for evaluation.
Hardware Specification No The paper mentions 'on a personal computer' but does not provide any specific hardware details such as GPU/CPU models or memory specifications.
Software Dependencies No The paper does not list any specific software dependencies with version numbers.
Experiment Setup Yes In PCG, we set η = 0.33 and use the closed-form expression for the multilinear relaxation of f D. We set ε = 0.1, δ = 1/m1.5 where m = |D| = 100, with which the privacy parameter used in the differentially private choices of increment is ε0 0.01006.