Summary of Expandable Subspace Ensemble For Pre-trained Model-based Class-incremental Learning, by Da-wei Zhou et al.
Expandable Subspace Ensemble for Pre-Trained Model-Based Class-Incremental Learning
by Da-Wei Zhou, Hai-Long Sun, Han-Jia Ye, De-Chuan Zhan
First submitted to arxiv on: 18 Mar 2024
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 addresses the challenge of Class-Incremental Learning (CIL) for Pre-Trained Models (PTMs), where new classes are learned without forgetting previous ones. The authors propose ExpAndable Subspace Ensemble (EASE) to efficiently update PTM-based models without harming former knowledge. EASE trains lightweight adapter modules for each new task, creating task-specific subspaces that enable joint decision-making. The approach also includes a semantic-guided prototype complement strategy to synthesize old classes’ new features without using old instances. Experimental results on seven benchmark datasets verify EASE’s state-of-the-art performance. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper is about helping machines learn new things without forgetting what they already know. It proposes a way to make big language models adapt to new tasks and still remember the old ones. The method uses special “adapters” that help the model focus on the new task, while keeping the old information. This approach works really well, according to tests on seven different datasets. |
