Summary of Learning From Noisy Labels For Long-tailed Data Via Optimal Transport, by Mengting Li et al.
Learning from Noisy Labels for Long-tailed Data via Optimal Transport
by Mengting Li, Chuang Zhu
First submitted to arxiv on: 7 Aug 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 This paper addresses a significant issue in deep learning model training: noisy labels in real-world datasets can impair model performance. The authors propose a novel approach to manage data with both long-tailed distributions and noisy labels. A loss-distance cross-selection module filters clean samples, addressing uncertainties introduced by noisy labels and long-tailed distributions. Optimal transport strategies generate pseudo-labels for the noise set in a semi-supervised training manner, enhancing pseudo-label quality while mitigating sample scarcity caused by the long-tailed distribution. The authors conduct experiments on synthetic and real-world datasets, demonstrating that their method surpasses current state-of-the-art methods. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper helps fix a big problem with deep learning models: when they’re trained on noisy data, it makes them work poorly. The researchers found a way to make this work better by using two new techniques. First, they created a special filter that finds the cleanest samples and ignores the noisy ones. Then, they used another trick called optimal transport to create fake labels for the noisy data, making it easier for the model to learn from. They tested their method on both pretend and real-world datasets and showed that it works better than what’s currently available. |
Keywords
» Artificial intelligence » Deep learning » Semi supervised
