Summary of Robust Subgraph Learning by Monitoring Early Training Representations, By Sepideh Neshatfar et al.
Robust Subgraph Learning by Monitoring Early Training Representations
by Sepideh Neshatfar, Salimeh Yasaei Sekeh
First submitted to arxiv on: 14 Mar 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Cryptography and Security (cs.CR)
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 introduces a novel technique called SHERD (Subgraph Learning Hale through Early Training Representation Distances) to enhance the performance and adversarial robustness of graph neural networks (GNNs). SHERD detects susceptible nodes during adversarial attacks using standard distance metrics and removes them to form a robust subgraph, maintaining node classification performance. The method is tested on multiple datasets, including citation networks and microanatomical tissue structures, showing substantial improvement in robust performance and outperforming several baselines. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary SHERD is a new way to make graph neural networks better at learning from graphs and more secure against attacks that try to trick them. It works by finding the “bad” nodes in a graph that might be used to attack it, and then removing those nodes to create a safer version of the graph. This makes the network more reliable and accurate, even when faced with tricky data. The technique is tested on different types of graphs, including ones related to science papers and the structure of cells in the placenta. |
Keywords
* Artificial intelligence * Classification
