Summary of Stochastic Bandits Robust to Adversarial Attacks, by Xuchuang Wang et al.
Stochastic Bandits Robust to Adversarial Attacks
by Xuchuang Wang, Jinhang Zuo, Xutong Liu, John C.S. Lui, Mohammad Hajiesmaili
First submitted to arxiv on: 16 Aug 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: None
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary Medium Difficulty summary: This research paper investigates stochastic multi-armed bandit algorithms that are resilient to adversarial attacks, where an attacker can observe the learner’s action and then alter their reward observation. The study focuses on two cases: with or without knowledge of an attack budget C, which is defined as an upper bound of the summation of the difference between actual and altered rewards. For both scenarios, the authors develop two types of algorithms with regret bounds that incorporate additive or multiplicative C dependence terms. Notably, for the known attack budget case, the proposed algorithms achieve regret bounds of O((K/Δ)log T + KC) and Õ(√KTC) for additive and multiplicative C terms, respectively. The authors also provide lower bound results demonstrating the tightness of their upper bounds and the optimality of their algorithms. This work highlights a fundamental distinction between bandits with attacks and corruption models. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Low Difficulty summary: Imagine you’re trying to find the best way to make money by choosing which of many investments to use. But, what if someone is secretly trying to manipulate your decision-making? That’s the problem this paper solves! The researchers created new algorithms that can adapt to these tricky situations where someone might try to change the outcome after you’ve made a choice. They tested their ideas in two different scenarios: when they knew how much the attacker could alter the rewards, and when they didn’t know. Their results show that their algorithms are really good at making smart choices even with these sneaky attacks happening. |
