Summary of Boosting Consistency in Dual Training For Long-tailed Semi-supervised Learning, by Kai Gan et al.
Boosting Consistency in Dual Training for Long-Tailed Semi-Supervised Learning
by Kai Gan, Tong Wei, Min-Ling Zhang
First submitted to arxiv on: 19 Jun 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 In this research paper, the authors propose a novel method called Boosting cOnsistency in duAl Training (BOAT) for long-tailed semi-supervised learning (LTSSL). LTSSL has gained popularity in various real-world classification problems, but existing algorithms often assume similar class distributions between labeled and unlabeled data. BOAT addresses this limitation by constructing two branches: a standard branch to improve head classes’ performance and a balanced branch to enhance tail classes’ performance. The model’s simplicity belies its effectiveness, as it achieves state-of-the-art performance on LTSSL benchmarks, with an average 2.7% increase in test accuracy when class distributions are mismatched. BOAT also outperforms many sophisticated algorithms even when class distributions are identical. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary BOAT is a simple and effective way to use unlabeled data for long-tailed semi-supervised learning. The method works by creating two branches that focus on different types of data: one branch looks at the most common classes (head) and another branch looks at less common classes (tail). By balancing these two branches, BOAT is able to improve its performance on both head and tail classes. This makes it a great tool for classification problems where there are many more examples of some classes than others. |
Keywords
» Artificial intelligence » Boosting » Classification » Semi supervised
