Summary of Uncovering Capabilities Of Model Pruning in Graph Contrastive Learning, by Junran Wu et al.
Uncovering Capabilities of Model Pruning in Graph Contrastive Learning
by Junran Wu, Xueyuan Chen, Shangzhe Li
First submitted to arxiv on: 27 Oct 2024
Categories
- Main: Machine Learning (cs.LG)
- 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 pre-training graph neural networks without ground-truth labels using graph contrastive learning. The traditional method of generating augmented views is criticized for altering semantics and undermining generalization. Instead, the authors suggest contrasting different model versions, rather than views. They theoretically prove that pruning models leads to better representation ability compared to data augmentations. In practice, they generate a perturbed graph encoder by pruning transformation weights and develop a local contrastive loss to tackle hard negative samples. The method is tested on various benchmarks for graph classification via unsupervised and transfer learning, outperforming state-of-the-art works. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper tries to solve a problem in machine learning called graph contrastive learning. This means teaching machines to understand graphs without having them labeled beforehand. The usual way of doing this is by creating many different versions of the same graph, but this changes what the graph means. The authors thought about this and came up with a new idea: instead of changing the graph, change how the machine sees it. They show that this works better than the old way and can even be used to learn from other graphs. This could be important for things like recognizing molecules or understanding social networks. |
Keywords
» Artificial intelligence » Classification » Contrastive loss » Encoder » Generalization » Machine learning » Pruning » Semantics » Transfer learning » Unsupervised
