Shuffle Private Linear Contextual Bandits

Authors: Sayak Ray Chowdhury, Xingyu Zhou

ICML 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We further perform simulations on synthetic data that corroborate our theoretical results. In this section, we empirically evaluate the regret performance of Algorithm 1 (under shuffle model), which we abbreviate as Lin UCB-SDP-Amp and Lin UCB-SDP-Vec when instantiated with PAmp and PVec, respectively.
Researcher Affiliation Academia 1Boston University, USA 2Wayne State University, USA. Correspondence to: Sayak Ray Chowdhury <sayak@bu.edu>, Xingyu Zhou <xingyu.zhou@wayne.edu>.
Pseudocode Yes Algorithm 1 Shuffle Private Lin UCB
Open Source Code Yes Code is available at https://github.com/sayakrc/Differentially-Private-Bandits.
Open Datasets No We further perform simulations on synthetic data that corroborate our theoretical results. The paper mentions "synthetic data" but does not provide any access information or citation for a public dataset.
Dataset Splits No For all the experiments, we consider 100 arms, set T = 20000 rounds, and average our results over 50 randomly generated bandit instances. Each instance is characterized by an (unknown) parameter θ and feature vectors of dimension d = 5. The paper describes the simulation parameters but does not specify train/validation/test dataset splits.
Hardware Specification No The paper does not provide any specific hardware details such as GPU/CPU models or types of machines used for running the experiments.
Software Dependencies No The paper does not provide specific ancillary software details with version numbers.
Experiment Setup Yes For all the experiments, we consider 100 arms, set T = 20000 rounds, and average our results over 50 randomly generated bandit instances. Each instance is characterized by an (unknown) parameter θ and feature vectors of dimension d = 5. ... We fix δ =0.1 and plot the results for varying privacy level ε {0.2, 1, 10}. We use Batchsize B = 20 for Lin UCB-SDP.