Summary of Learning Label Refinement and Threshold Adjustment For Imbalanced Semi-supervised Learning, by Zeju Li et al.
Learning Label Refinement and Threshold Adjustment for Imbalanced Semi-Supervised Learning
by Zeju Li, Ying-Qiu Zheng, Chen Chen, Saad Jbabdi
First submitted to arxiv on: 7 Jul 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 Semi-supervised learning (SSL) algorithms face challenges when dealing with imbalanced training data. Pseudo-labeling, a key component of SSL, can amplify biases towards the majority class. This paper investigates pseudo-labeling strategies for imbalanced SSL, including refinement and threshold adjustment, through statistical analysis. The authors find that existing SSL algorithms relying on heuristic pseudo-label generation or uncalibrated model confidence are unreliable when faced with imbalanced data. To address this, they introduce SEVAL (SEmi-supervised learning with pseudo-label optimization based on VALidation data), which optimizes refinement and thresholding parameters using a class-balanced partition of the training dataset. SEVAL adapts to specific tasks and delivers more accurate and effective pseudo-labels in various imbalanced SSL scenarios. Experimental results show that SEVAL outperforms state-of-the-art SSL methods, demonstrating its potential to enhance various SSL techniques. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper looks at how machine learning algorithms called semi-supervised learning (SSL) work when the data is not balanced. Imagine you’re trying to teach a dog new tricks, but most of the time it gets them right, and only sometimes gets them wrong. The algorithm would learn to rely on its “right” guesses too much and ignore the rare mistakes. This paper proposes a solution called SEVAL (SEmi-supervised learning with pseudo-label optimization based on VALidation data), which helps the algorithm focus more on the minority class and less on the majority one. By testing SEVAL, the authors found that it outperforms existing algorithms in these situations. |
Keywords
» Artificial intelligence » Machine learning » Optimization » Semi supervised
