Summary of Empirical Perturbation Analysis Of Linear System Solvers From a Data Poisoning Perspective, by Yixin Liu et al.
Empirical Perturbation Analysis of Linear System Solvers from a Data Poisoning Perspective
by Yixin Liu, Arielle Carr, Lichao Sun
First submitted to arxiv on: 1 Oct 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Cryptography and Security (cs.CR); Numerical Analysis (math.NA)
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 The proposed research aims to analyze linear solvers’ responses to data poisoning attacks in machine learning settings. Specifically, it investigates how errors in input data affect the accuracy and fitting error of solution from a linear system-solving algorithm under perturbations common in adversarial attacks. The study proposes two methods of perturbation: Label-guided Perturbation (LP) and Unconditioning Perturbation (UP), and identifies which solvers are most impacted by different types of attacks. By reframing the analysis through the lens of data poisoning, this work contributes to developing more robust linear solvers and provides insights into poisoning attacks on linear solvers. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper looks at how machine learning models can be attacked by intentionally changing the input data. The researchers want to know which types of models are most affected by these attacks. They propose two new ways to make changes to the data, called Label-guided Perturbation (LP) and Unconditioning Perturbation (UP). The goal is to make machine learning models more robust against these kinds of attacks. |
Keywords
* Artificial intelligence * Machine learning
