Optimizing Graph Neural Networks for Jet Tagging in Particle Physics on FPGAs

Optimizing Graph Neural Networks for Jet Tagging in Particle Physics on FPGAs
复制标题

优化 FPGA 上粒子物理中喷射标记的图神经网络

DOI:
10.1109/fpl57034.2022.00057
复制
发表时间:
2022
期刊:
2022 32nd International Conference on Field-Programmable Logic and Applications (FPL)
影响因子:
--
通讯作者:
W. Luk
W. Luk
中科院分区:
--
文献类型:
--
作者:
Zhiqiang Que;Marcus Loo;Hongxiang Fan;M. Pierini;A. Tapper;W. Luk

文献摘要

参考文献

被引文献

相似文献

这项工作提出了一种新的可重构架构,以减少延迟的JEDI网络,图形神经网络(GNN)为基础的算法,在粒子物理学中的射流标记,达到了最先进的精度。加速JEDI-net具有挑战性,因为它需要低延迟来部署CERN大型强子对撞机的事件选择网络。针对基于GNN的JEDI网,提出了一种基于外积的矩阵乘法方法,提高了数据空间局部性,减少了设计延迟。它通过强度降低的代码转换进一步增强,该代码转换利用稀疏模式和二进制邻接矩阵来提高硬件效率,同时减少延迟。此外,该架构的可定制模板已经设计并开源,这使得能够使用高级综合工具生成低延迟FPGA设计,并有效利用资源。评估结果表明,我们的FPGA实现的速度高达9.5倍,消耗高达6.5倍,比GPU实现的功耗。此外,我们的FPGA设计的吞吐量足够高,可以在亚微秒、实时对撞机触发系统中部署JEDI-net,使其能够从更高的准确性中受益。
This work proposes a novel reconfigurable architecture for reducing the latency of JEDI-net, a Graph Neural Network (GNN) based algorithm for jet tagging in particle physics, which achieves state-of-the-art accuracy. Accelerating JEDI-net is challenging since it requires low latency to deploy the network for event selection at the CERN Large Hadron Collider. This paper proposes an outer-product based matrix multiplication approach customized for GNN-based JEDI-net, which increases data spatial locality and reduces design latency. It is further enhanced by code transformation with strength reduction which exploits sparsity patterns and binary adjacency matrices to increase hardware efficiency while reducing latency. In addition, a customizable template for this architecture has been designed and open-sourced, which enables the generation of low-latency FPGA designs with efficient resource utilization using high-level synthesis tools. Evaluation results show that our FPGA implementation is up to 9.5 times faster and consumes up to 6.5 times less power than a GPU implementation. Moreover, the throughput of our FPGA design is sufficiently high to enable deployment of JEDI-net in a sub-microsecond, real-time collider trigger system, enabling it to benefit from improved accuracy.
DOI: 10.48550/arxiv.2203.05095
发表时间: 2022-03
期刊: 2022 IEEE International Parallel and Distributed Processing Symposium (IPDPS)
影响因子: --
作者:
Hongkuan Zhou;Bingyi Zhang;R. Kannan;V. Prasanna;Carl E. Busart
通讯作者: Hongkuan Zhou;Bingyi Zhang;R. Kannan;V. Prasanna;Carl E. Busart
DOI: 10.1140/epjc/s10052-021-09675-8
发表时间: 2021-03
期刊: The European Physical Journal C
影响因子: --
作者:
X. Ju;D. Murnane;P. Calafiura;Nicholas Choma;S. Conlon;S. Farrell;Yaoyuan Xu;M. Spiropulu;J. Vlimant;A. Aurisano;J. Hewes;G. Cerati;L. Gray;T. Klijnsma;J. Kowalkowski;M. Atkinson;M. Neubauer;G. Dezoort;S. Thais;Aditi Chauhan;A. Schuy;Shih-Chieh Hsu;A. Ballow;A. Lazar
通讯作者: X. Ju;D. Murnane;P. Calafiura;Nicholas Choma;S. Conlon;S. Farrell;Yaoyuan Xu;M. Spiropulu;J. Vlimant;A. Aurisano;J. Hewes;G. Cerati;L. Gray;T. Klijnsma;J. Kowalkowski;M. Atkinson;M. Neubauer;G. Dezoort;S. Thais;Aditi Chauhan;A. Schuy;Shih-Chieh Hsu;A. Ballow;A. Lazar
DOI: 10.1145/3373087.3375312
发表时间: 2019-12
期刊: Proceedings of the 2020 ACM/SIGDA International Symposium on Field-Programmable Gate Arrays
影响因子: --
作者:
Hanqing Zeng;V. Prasanna
通讯作者: Hanqing Zeng;V. Prasanna
使用高级综合的 GCN 推理加速
DOI: 10.1109/hpec49654.2021.9622801
发表时间: 2021
期刊: 2021.
影响因子: --
作者:
Lin, Yi Chien;Zhang, Bingyi;Prasanna, Viktor
通讯作者: Prasanna, Viktor
DOI: 10.1109/fccm51124.2021.00012
发表时间: 2021-02
期刊: 2021 IEEE 29th Annual International Symposium on Field-Programmable Custom Computing Machines (FCCM)
影响因子: --
作者:
Bingyi Zhang;R. Kannan;V. Prasanna
通讯作者: Bingyi Zhang;R. Kannan;V. Prasanna