Summary of Towards Realistic Long-tailed Semi-supervised Learning in An Open World, by Yuanpeng He et al.
Towards Realistic Long-tailed Semi-supervised Learning in an Open World
by Yuanpeng He, Lijian Li
First submitted to arxiv on: 23 May 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- 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 proposed Open-world Long-tailed Semi-supervised Learning (OLSSL) algorithm, dubbed ROLSSL, tackles the limitations of existing OLSSL methods by assuming no distribution relationships between known and novel categories. This setting is more realistic, as it acknowledges that acquiring valid annotations can be costly in real-world scenarios. The solution involves a dual-stage post-hoc logit adjustment strategy that considers sample frequency, total category count, and data size to dynamically adjust predictive probabilities and mitigate category bias. Experimental results on CIFAR100 and ImageNet100 demonstrate performance improvements of up to 50.1%, establishing the proposed method as a strong baseline for this task. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Imagine you have a special kind of learning problem where you need to identify things in pictures, but sometimes those pictures are new and never seen before. The problem is that these new pictures might look very different from the old ones. To solve this, researchers created a new way to learn called Open-world Long-tailed Semi-supervised Learning (OLSSL). This method tries to figure out how to tell apart the old pictures from the new ones, even if they’re not similar at all. The idea is that by looking at many pictures and seeing which ones are more common or less common, we can learn how to identify the new pictures better. This approach worked really well on some famous image recognition datasets, showing that it’s a good way to solve this kind of problem. |
Keywords
» Artificial intelligence » Semi supervised
