Summary of From Overfitting to Robustness: Quantity, Quality, and Variety Oriented Negative Sample Selection in Graph Contrastive Learning, by Adnan Ali et al.
From Overfitting to Robustness: Quantity, Quality, and Variety Oriented Negative Sample Selection in Graph Contrastive Learning
by Adnan Ali, Jinlong Li, Huanhuan Chen, Ali Kashif Bashir
First submitted to arxiv on: 21 Jun 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 proposed Cumulative Sample Selection (CSS) algorithm addresses overfitting in Graph Contrastive Learning (GCL) by comprehensively considering negative sample quality, variation, and quantity. This study presents a novel framework called NegAmplify that integrates CSS into GCL. The algorithm constructs three negative sample pools based on ease of classification and selects 10% samples from each pool for training. A decision agent module evaluates model performance and decides whether to explore more negative samples or maintain the current ratio. NegAmplify is compared with state-of-the-art methods on nine graph node classification datasets, achieving better accuracy in seven datasets with up to 2.86% improvement. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This study improves Graph Contrastive Learning (GCL) by solving overfitting problems. GCL compares positive and negative samples to learn about nodes. But if there’s too little variety or bad quality in these negative samples, the model only works well for specific nodes. To fix this, researchers developed a new way to choose negative samples called Cumulative Sample Selection (CSS). CSS looks at how hard it is to classify negative samples and picks the best ones. Then, an agent helps decide if they should keep trying or stop. This new framework, NegAmplify, was tested on many datasets and did better than other top methods in some cases. |
Keywords
» Artificial intelligence » Classification » Overfitting
