Summary of Adaptgcd: Multi-expert Adapter Tuning For Generalized Category Discovery, by Yuxun Qu et al.
AdaptGCD: Multi-Expert Adapter Tuning for Generalized Category Discovery
by Yuxun Qu, Yongqiang Tang, Chenyang Zhang, Wensheng Zhang
First submitted to arxiv on: 29 Oct 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 paper proposes a novel approach to Generalized Category Discovery (GCD), which aims to classify old categories while also discovering new ones in unlabeled data. Existing methods focus on transferring general knowledge from pre-trained models to GCD tasks, but these fine-tuning strategies fail to balance between generalization capacity and adaptability. The proposed adapter-tuning-based method, AdaptGCD, introduces a multi-expert adapter structure that separates old and new classes into different expert groups. This approach is shown to be effective through extensive experiments on 7 widely-used datasets. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper is about a new way to learn from data that’s not labeled or partially labeled. It’s like trying to figure out what new categories of things are in the world, while also recognizing old ones. Most people try to solve this problem by taking pre-trained models and adjusting them to fit this task, but it’s hard to get the balance right between recognizing old things and discovering new ones. The authors propose a new way to do this called AdaptGCD, which uses multiple experts to help separate old and new categories. This approach works well on lots of different datasets. |
Keywords
» Artificial intelligence » Fine tuning » Generalization
