Adversarial Attacks on Stochastic Bandits

Authors: Kwang-Sung Jun, Lihong Li, Yuzhe Ma, Jerry Zhu

NeurIPS 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In this section, we run simulations on attacking -greedy and UCB algorithms to illustrate our theoretical findings.
Researcher Affiliation Collaboration Kwang-Sung Jun Boston University kwangsung.jun@gmail.com Lihong Li Google Brain lihong@google.com Yuzhe Ma UW-Madison ma234@wisc.edu Xiaojin Zhu UW-Madison jerryzhu@cs.wisc.edu
Pseudocode Yes Algorithm 1 Alice s attack against a bandit algorithm
Open Source Code No No explicit statement or link for open-source code for the methodology.
Open Datasets No The reward distributions of arms 1 and 2 are N(µ1, σ2) and N(0, σ2), respectively, with µ1 > 0.
Dataset Splits No We repeat 1000 trials.
Hardware Specification No No specific hardware details are mentioned.
Software Dependencies No No specific software dependencies with version numbers are mentioned.
Experiment Setup Yes We let δ = 0.025. Bob s exploration probability decays as t = 1 t . We run Alice and Bob for T = 105 rounds; this forms one trial. We repeat 1000 trials. ... We let δ = 0.05. ...We perform 100 trials with T = 107 rounds.