Summary of Imbagcd: Imbalanced Generalized Category Discovery, by Ziyun Li et al.
ImbaGCD: Imbalanced Generalized Category Discovery
by Ziyun Li, Ben Dai, Furkan Simsek, Christoph Meinel, Haojin Yang
First submitted to arxiv on: 4 Dec 2023
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Machine Learning (cs.LG)
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 This paper proposes a novel framework for Generalized Class Discovery (GCD) that addresses the issue of imbalanced class distributions in unlabeled datasets. Unlike previous approaches that assume equal frequencies for known and unknown categories, the proposed method, called ImbaGCD, takes into account the long-tailed property of visual classes, where common classes are more frequent than rare ones. The framework uses optimal transport-based expectation maximization to align the marginal class prior distribution and incorporates a systematic mechanism for estimating the imbalanced class prior distribution. Experimental results on CIFAR-100 and ImageNet-100 show that ImbaGCD outperforms previous state-of-the-art GCD methods, achieving an improvement of approximately 2-4% on CIFAR-100 and 15-19% on ImageNet-100. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Imbalanced Generalized Category Discovery is a new challenge in machine learning. Normally, we try to find categories (or classes) in data that hasn’t been labeled yet. But what if some categories are much more common than others? This paper solves this problem by creating a new way of doing category discovery that takes into account how often each category appears. |
Keywords
* Artificial intelligence * Machine learning
