Summary of On the Effectiveness Of Supervision in Asymmetric Non-contrastive Learning, by Jeongheon Oh et al.
On the Effectiveness of Supervision in Asymmetric Non-Contrastive Learning
by Jeongheon Oh, Kibok Lee
First submitted to arxiv on: 16 Jun 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (stat.ML)
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 Supervised contrastive representation learning has been successful in various transfer learning situations, but the extension of asymmetric non-contrastive learning (ANCL) to supervised scenarios is less explored. This study bridges the gap by proposing SupSiam and SupBYOL, which leverage labels to improve representation learning while avoiding collapse. The framework improves representation quality while reducing intra-class variance. Adjusting the contribution of supervision leads to better performance. Experimental results demonstrate the superiority of supervised ANCL across various datasets and tasks. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Supervised contrastive representation learning is a way to make computer systems smarter by teaching them to recognize patterns in data. In this study, researchers tried to make a type of learning called asymmetric non-contrastive learning work better when we have labels (information) to help the system learn. They came up with two new ways to do this, which they call SupSiam and SupBYOL. These methods use the labels to teach the system to recognize patterns in data better. The researchers found that these methods make the system learn faster and more accurately than before. This is important because it can help us make computer systems that are smarter and more helpful. |
Keywords
» Artificial intelligence » Representation learning » Supervised » Transfer learning
