Summary of Less Is More: Hop-wise Graph Attention For Scalable and Generalizable Learning on Circuits, by Chenhui Deng et al.
Less is More: Hop-Wise Graph Attention for Scalable and Generalizable Learning on Circuits
by Chenhui Deng, Zichao Yue, Cunxi Yu, Gokce Sarar, Ryan Carey, Rajeev Jain, Zhiru Zhang
First submitted to arxiv on: 2 Mar 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Hardware Architecture (cs.AR)
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 This paper proposes a novel attention-based model, called HOGA, for learning scalable and generalizable circuit representations. Unlike graph neural networks (GNNs), which struggle with large graphs and limited generalizability, HOGA computes hop-wise features per node before training, then uses gated self-attention to learn important features across different hops. This approach enables HOGA to adapt to various structures across circuits and be trained efficiently in a distributed manner. The paper demonstrates the efficacy of HOGA for quality of results (QoR) prediction and functional reasoning in electronic design automation (EDA) tasks, achieving significant improvements over conventional GNNs. Specifically, HOGA reduces QoR estimation error by 46.76% after logic synthesis and improves functional block identification accuracy by 10.0% on unseen gate-level netlists after complex technology mapping. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This research paper introduces a new way to understand electronic circuits. The current approach, called graph neural networks (GNNs), has some limitations when dealing with very large circuits. To solve this problem, the authors propose a new model called HOGA, which is more efficient and can handle different types of circuit structures. HOGA works by looking at each part of the circuit separately and then combining that information to get a better understanding of the whole circuit. The paper shows that HOGA performs better than GNNs in predicting how well a circuit will work and in identifying what parts of the circuit are responsible for certain functions. |
Keywords
* Artificial intelligence * Attention * Self attention
