Summary of Do Generated Data Always Help Contrastive Learning?, by Yifei Wang et al.
Do Generated Data Always Help Contrastive Learning?
by Yifei Wang, Jizhe Zhang, Yisen Wang
First submitted to arxiv on: 19 Mar 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 In this paper, researchers investigate the impact of generative models on contrastive learning for unsupervised visual representation. They find that while generated high-quality images can enhance contrastive representation learning, they can also sometimes harm it. The authors reveal the complementary roles of data inflation and augmentation, showing that stronger inflation requires weaker augmentations, and vice versa. To address this, they propose Adaptive Inflation (AdaInf), a data-centric strategy without extra computation cost. AdaInf brings significant improvements to various contrastive learning methods on benchmark datasets, setting a new record on CIFAR-10 with SimCLR. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Contrastive Learning (CL) helps computers learn without labels by comparing similar and dissimilar images. This paper looks at how generated fake images can help or hurt CL. The researchers found that while these fake images can make CL better in some cases, they can also make it worse if not used carefully. To solve this problem, they came up with a new way to use the fake images called Adaptive Inflation (AdaInf). This method works well and even sets a new record for performance on certain tasks. |
Keywords
* Artificial intelligence * Representation learning * Unsupervised
