Summary of Robust Testing For Deep Learning Using Human Label Noise, by Gordon Lim et al.
Robust Testing for Deep Learning using Human Label Noise
by Gordon Lim, Stefan Larson, Kevin Leach
First submitted to arxiv on: 29 Nov 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 The proposed paper addresses the issue of label noise in training datasets for deep learning systems. It highlights that current methods for testing Learning with Noisy Labels (LNL) use synthetic label noise, which may not accurately reflect the type of noise introduced by human labeling. To address this gap, the authors introduce Cluster-Based Noise (CBN), a method for generating feature-dependent noise that simulates human-like label noise. The paper demonstrates that current LNL methods perform worse when tested using CBN, making it a more rigorous approach to testing neural networks. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper is about how computers learn from noisy labels. Sometimes, people make mistakes when labeling data, which can make the computer learn incorrectly. Researchers have developed ways to train computers to work well even with noisy labels, but they usually test these methods using fake noise. The problem is that this doesn’t reflect real-world mistakes people might make. To fix this, the authors created a new way to generate noisy labels that mimic human errors. This helps evaluate how well current methods can handle real-world noise. The results show that these methods need improvement. |
Keywords
» Artificial intelligence » Deep learning
