Summary of Addressing Long-tail Noisy Label Learning Problems: a Two-stage Solution with Label Refurbishment Considering Label Rarity, by Ying-hsuan Wu et al.
Addressing Long-Tail Noisy Label Learning Problems: a Two-Stage Solution with Label Refurbishment Considering Label Rarity
by Ying-Hsuan Wu, Jun-Wei Hsieh, Li Xin, Shin-You Teng, Yi-Kuan Hsieh, Ming-Ching Chang
First submitted to arxiv on: 4 Mar 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 The paper proposes a novel approach to address the issue of noisy labels and class imbalance in real-world datasets, particularly those exhibiting long-tailed distributions. The authors introduce a two-stage method combining soft-label refurbishing with multi-expert ensemble learning. In the first stage, they use contrastive learning to obtain unbiased features and make preliminary predictions using a classifier trained with a balanced noise-tolerant cross-entropy (BANC) loss. The second stage involves applying label refurbishment methods to obtain soft labels for multi-expert ensemble learning. This principled solution is designed to tackle the long-tail noisy label problem. Experiments on multiple benchmarks, including CIFAR-10 and CIFAR-100, as well as real-noise datasets like Food-101N and Animal-10N, demonstrate the superiority of this approach, achieving state-of-the-art accuracies. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper solves a big problem in machine learning: when you have data with mistakes in it. This happens often in real-world datasets where some classes are more common than others. The authors come up with a new way to fix these errors by first getting better features and then using multiple experts to make predictions. They test this method on several datasets and show that it works much better than existing methods. |
Keywords
* Artificial intelligence * Cross entropy * Machine learning
