Summary of Manipulating Predictions Over Discrete Inputs in Machine Teaching, by Xiaodong Wu et al.
Manipulating Predictions over Discrete Inputs in Machine Teaching
by Xiaodong Wu, Yufei Han, Hayssam Dahrouj, Jianbing Ni, Zhenwen Liang, Xiangliang Zhang
First submitted to arxiv on: 31 Jan 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI)
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 In this paper, researchers explore machine teaching in the discrete domain, where they aim to manipulate a student model’s predictions based on a teacher’s goals. They formulate this task as a combinatorial optimization problem and propose an iterative searching algorithm to solve it. The algorithm demonstrates significant merit in scenarios where a teacher corrects erroneous predictions or maliciously manipulates the model to misclassify specific samples. Experimental results show that the proposed algorithm outperforms conventional baselines in effectively and efficiently manipulating the model’s predictions. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Machine teaching helps teachers create optimal datasets for student models to achieve specific goals. Researchers have studied this topic in continuous domains, but there is limited work on discrete domains. This paper looks at how to manipulate student models’ predictions based on teacher goals by changing training data efficiently. They use an iterative searching algorithm and show that it can correct errors or misclassify samples maliciously. The results are superior to conventional methods. |
Keywords
* Artificial intelligence * Optimization * Student model
