arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2501.13610cs.NEcs.AI
收藏

数字事件驱动AI加速器中的高效突触延迟实现

Efficient Synaptic Delay Implementation in Digital Event-Driven AI Accelerators

Instituto de Microelectrónica de Sevilla, CSIC/Universidad de Sevilla(塞维利亚微电子研究所,西班牙国家研究委员会/塞维利亚大学);Eindhoven University of Technology(埃因霍温理工大学);IMEC-Netherlands(比利时微电子研究中心荷兰分部);University of Twente(特文特大学)
Roy Meijer, Paul Detterer, Amirreza Yousefzadeh, Alberto Patino-Saucedo, Guanghzi Tang, Kanishkan Vadivel, Yinfu Xu, Manil-Dev Gomony, Federico Corradi, Bernabe… 展开作者

Roy Meijer, Paul Detterer, Amirreza Yousefzadeh, Alberto Patino-Saucedo, Guanghzi Tang, Kanishkan Vadivel, Yinfu Xu, Manil-Dev Gomony, Federico Corradi, Bernabe Linares-Barranco, Manolis Sifalakis

AI总结:

针对神经网络突触延迟参数化未被充分探索的问题,提出共享循环延迟队列(SCDQ)硬件结构,用于数字神经形态加速器支持突触延迟,在内存扩展性、算法-硬件协同优化及延迟、面积和能耗方面表现更优。

摘要:

神经网络模型的突触延迟参数化在很大程度上仍未得到探索,但最近的文献已显示出有前景的结果,表明延迟参数化模型比性能相当(例如任务准确率)的非延迟参数化模型更简单、更小、更稀疏,因此更节能。我们提出了共享循环延迟队列(SCDQ),一种用于在数字神经形态加速器上支持突触延迟的新型硬件结构。我们的分析和硬件结果表明,在内存方面,它比当前常用的方法具有更好的扩展性,并且更易于进行算法-硬件协同优化,实际上,内存扩展受模型稀疏性而非仅仅是网络规模的调节。除了内存,我们还报告了延迟、面积和每次推理能耗方面的性能。

英文摘要:

Synaptic delay parameterization of neural network models have remained largely unexplored but recent literature has been showing promising results, suggesting the delay parameterized models are simpler, smaller, sparser, and thus more energy efficient than similar performing (e.g. task accuracy) non-delay parameterized ones. We introduce Shared Circular Delay Queue (SCDQ), a novel hardware structure for supporting synaptic delays on digital neuromorphic accelerators. Our analysis and hardware results show that it scales better in terms of memory, than current commonly used approaches, and is more amortizable to algorithm-hardware co-optimizations, where in fact, memory scaling is modulated by model sparsity and not merely network size. Next to memory we also report performance on latency area and energy per inference.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑