Summary of Graph Edit Distance with General Costs Using Neural Set Divergence, by Eeshaan Jain et al.
Graph Edit Distance with General Costs Using Neural Set Divergence
by Eeshaan Jain, Indradyumna Roy, Saswat Meher, Soumen Chakrabarti, Abir De
First submitted to arxiv on: 26 Sep 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 A novel neural network-based approach called GRAPHEDX is proposed to estimate the Graph Edit Distance (GED) between two given graphs. Unlike existing methods, GRAPHEDX can handle edit operations with different costs by representing each graph as a set of node and edge embeddings and using these representations to design a family of neural set divergence surrogates. The GED computation problem is reformulated as a quadratic assignment problem (QAP), which is then solved using a Gumbel-Sinkhorn permutation generator that learns alignments between nodes and edges, ensuring consistency with each other. Experiments on various datasets demonstrate that GRAPHEDX outperforms state-of-the-art methods in terms of prediction error. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary GRAPHEDX is a new way to measure how similar or different two graphs are. Right now, there’s no easy way to do this because the problem is really hard to solve exactly. Some people have tried using special kinds of neural networks, but they didn’t work very well because they didn’t take into account that some edit operations might be more important than others. The new approach represents each graph as a set of tiny pieces and uses these pieces to create a way to compare the graphs. It also makes sure that the way it compares the nodes and edges is consistent with how the edges are connected. This method works really well on different kinds of datasets, even when the costs of different edit operations are different. |
Keywords
* Artificial intelligence * Neural network
