Automatic Routability Predictor Development Using Neural Architecture Search
Automatic Routability Predictor Development Using Neural Architecture Search
复制标题
DOI：
10.1109/iccad51958.2021.9643483
复制
发表时间：
2020-12
期刊：
影响因子：
--
通讯作者：
Jingyu Pan;Chen-Chia Chang;Tunhou Zhang;Zhiyao Xie;Jiang Hu;Weiyi Qi;Chung-Wei Lin;Rongjian Liang;Joydeep Mitra;Elias Fallon;Yiran Chen
中科院分区：
文献类型：
--
作者：
Jingyu Pan;Chen-Chia Chang;Tunhou Zhang;Zhiyao Xie;Jiang Hu;Weiyi Qi;Chung-Wei Lin;Rongjian Liang;Joydeep Mitra;Elias Fallon;Yiran Chen
机器学习技术的兴起激发了其在电子设计自动化（EDA）中的应用热潮，并有助于提高芯片设计的自动化程度。然而，手工制作的机器学习模型需要广泛的人类专业知识和巨大的工程努力。在这项工作中，我们利用神经架构搜索（NAS）来自动开发用于可布线性预测的高质量神经架构，这可以帮助引导单元布局走向可布线的解决方案。我们的搜索方法支持各种操作和高度灵活的连接，从而产生与之前所有人工模型显着不同的架构。在大型数据集上的实验结果表明，我们自动生成的神经架构明显优于多个有代表性的手工制作的解决方案。与手工构建的模型的最佳情况相比，NAS生成的模型在预测具有DRC违规的网络数量方面实现了5.85%的Kendall $T$，并且在DRC热点检测方面实现了2.12%的ROC曲线下面积（ROC-AUC）。此外，与人工制作的模型相比，这很容易需要几周的时间来开发，我们高效的NAS方法仅用0.3天就完成了整个自动搜索过程。
The rise of machine learning technology inspires a boom of its applications in electronic design automation (EDA) and helps improve the degree of automation in chip designs. However, manually crafted machine learning models require extensive human expertise and tremendous engineering efforts. In this work, we leverage neural architecture search (NAS) to automate the development of high-quality neural architectures for routability prediction, which can help to guide cell placement toward routable solutions. Our search method supports various operations and highly flexible connections, leading to architectures significantly different from all previous human-crafted models. Experimental results on a large dataset demonstrate that our automatically generated neural architectures clearly outperform multiple representative manually crafted solutions. Compared to the best case of manually crafted models, NAS-generated models achieve 5.85% higher Kendall's $T$ in predicting the number of nets with DRC violations and 2.12% better area under ROC curve (ROC-AUC) in DRC hotspot detection. Moreover, compared with human-crafted models, which easily take weeks to develop, our efficient NAS approach finishes the whole automatic search process with only 0.3 days.