Summary of Nc-ncd: Novel Class Discovery For Node Classification, by Yue Hou et al.
NC-NCD: Novel Class Discovery for Node Classification
by Yue Hou, Xueyuan Chen, He Zhu, Romei Liu, Bowen Shi, Jiaheng Liu, Junran Wu, Ke Xu
First submitted to arxiv on: 25 Jul 2024
Categories
- Main: Machine Learning (cs.LG)
- 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 novel Class Discovery (NCD) method for identifying new categories in unlabeled data faces a challenge in balancing the performance of old and new categories. Existing NCD approaches often struggle with catastrophic forgetting or inability to learn new categories. To address this, we introduce a node classification-based NCD scenario (NC-NCD) and propose the SWORD framework, which combines self-training, prototype replay, and distillation to enable models to cluster unlabeled new category nodes while preserving performance on old categories. Our approach achieves this by employing a self-training strategy to learn new categories and preventing forgetting of old categories through feature prototypes and knowledge distillation. Experimental results on four benchmarks demonstrate the superiority of SWORD over state-of-the-art methods. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary NCD is a way to find new categories in data without labels. The problem is that most NCD methods don’t do well when they have to remember old categories or learn new ones. This is because they tend to forget what they learned about the old categories or struggle to learn anything new. To fix this, we created a new way of doing NCD called NC-NCD and developed a special method called SWORD that helps models keep learning without forgetting what they already know. Our approach works by teaching models to learn new categories while keeping track of what they learned about the old ones. |
Keywords
» Artificial intelligence » Classification » Distillation » Knowledge distillation » Self training
