Summary of Incremental Self-training For Semi-supervised Learning, by Jifeng Guo et al.
Incremental Self-training for Semi-supervised Learning
by Jifeng Guo, Zhulin Liu, Tong Zhang, C. L. Philip Chen
First submitted to arxiv on: 14 Apr 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 enables machine learning models to function without relying heavily on labeled data. Self-training (ST) is an efficient technique that has gained attention for addressing noisy pseudo-labels. However, previous works have neglected the effective utilization of unlabeled data and the problem of high time consumption caused by iterative learning. This paper proposes Incremental Self-training (IST), a novel approach that processes data in batches, prioritizes pseudo-labeling on samples with high certainty, and then refines the classifier performance around the decision boundary after stabilization. IST is simple yet effective, fitting into existing self-training-based semi-supervised learning methods. The proposed method is evaluated on five datasets and two types of backbone, demonstrating improved recognition accuracy and learning speed, outperforming state-of-the-art competitors in three challenging image classification tasks. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper talks about a way to make machine learning models work better without needing lots of labeled data. They use something called self-training, which is an efficient method that helps with noisy labels. But there was a problem: it took too long to learn and didn’t use the unlabeled data efficiently. So, they came up with a new idea called Incremental Self-training (IST). IST does things in batches and focuses on samples where it’s pretty sure what the right answer is. Then, it fine-tunes its performance around the decision boundary. The result is that IST is simple yet effective and can be used with existing methods. They tested IST on five datasets and two types of models, showing that it works better and faster than some other approaches. |
Keywords
» Artificial intelligence » Attention » Image classification » Machine learning » Self training » Semi supervised
