Summary of Slca++: Unleash the Power Of Sequential Fine-tuning For Continual Learning with Pre-training, by Gengwei Zhang et al.
SLCA++: Unleash the Power of Sequential Fine-tuning for Continual Learning with Pre-training
by Gengwei Zhang, Liyuan Wang, Guoliang Kang, Ling Chen, Yunchao Wei
First submitted to arxiv on: 15 Aug 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI); 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 presents an in-depth analysis of the progressive overfitting problem in continual learning with pre-training (CLPT). The authors discuss how the use of strong pre-trained models (PTMs) can alleviate catastrophic forgetting, but also suffers from progressive overfitting. They argue that current efforts to address this issue by keeping PTMs frozen and incorporating task-specific prompts are sub-optimal. Instead, they propose a framework called Slow Learner with Classifier Alignment (SLCA++) that selectively reduces the learning rate of backbone parameters and aligns disjoint classification layers in a post-hoc fashion. The authors evaluate their approach on various continual learning scenarios and show that it outperforms state-of-the-art methods by a large margin. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper is about how computers can learn new things without forgetting what they already know. It’s called “continual learning” and it’s important for artificial intelligence to be good at this. The authors talk about how using pre-trained models can help, but also cause problems like “progressive overfitting”. They propose a new way of doing continual learning that works better than what people are currently doing. They tested their method on some computer vision tasks and it did really well. |
Keywords
» Artificial intelligence » Alignment » Classification » Continual learning » Overfitting
