Summary of Controlling Forgetting with Test-time Data in Continual Learning, by Vaibhav Singh et al.
Controlling Forgetting with Test-Time Data in Continual Learning
by Vaibhav Singh, Rahaf Aljundi, Eugene Belilovsky
First submitted to arxiv on: 19 Jun 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: None
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 In this paper, researchers tackle the challenge of updating foundational vision-language models to perform well on new tasks or domains without sacrificing their prior knowledge. They argue that test-time data can be used to refresh the model’s memory of previously learned tasks in a self-supervised manner, reducing forgetting at no extra labeling cost. The authors propose a simple student-teacher model with gradient-based sparse parameter updates and demonstrate significant performance improvements and reduction in forgetting. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper finds ways to update foundational vision-language models so they can perform well on new tasks or domains without losing their old knowledge. They think that test-time data can be used to help the model remember what it learned before, without needing any extra labeled data. The researchers suggest a simple way to do this using a student-teacher model and show that it works really well. |
Keywords
» Artificial intelligence » Self supervised » Teacher model
