Summary of Safe Semi-supervised Contrastive Learning Using In-distribution Data As Positive Examples, by Min Gu Kwak et al.
Safe Semi-Supervised Contrastive Learning Using In-Distribution Data as Positive Examples
by Min Gu Kwak, Hyungu Kahng, Seoung Bum Kim
First submitted to arxiv on: 3 Aug 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
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 proposed semi-supervised learning method applies a self-supervised contrastive learning approach to fully utilize unlabeled data, addressing class distribution mismatch scenarios where out-of-distribution (OOD) data exist. The method aggregates labeled negative examples of the same class into positive examples using a contrastive loss function with a coefficient schedule. Experimental results on image classification datasets – CIFAR-10, CIFAR-100, Tiny ImageNet, and CIFAR-100+Tiny ImageNet – under various mismatch ratios show significant improvements in classification accuracy. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Semi-supervised learning helps machines learn from a small amount of labeled data and a large amount of unlabeled data. The problem is that when there’s no equal class distribution, the method doesn’t work well. To solve this issue, scientists propose using self-supervised contrastive learning to make the most of all available data. This approach groups similar data points together, regardless of their classes. By doing so, it improves the accuracy of image classification tasks on datasets like CIFAR-10 and more. |
Keywords
* Artificial intelligence * Classification * Contrastive loss * Image classification * Self supervised * Semi supervised
