Resilience-Aware Frequency Tuning for Neural-Network-Based Approximate Computing Chips

被引:17
|
作者
Wang, Ying [1 ]
Deng, Jiachao [1 ]
Fang, Yuntan [1 ]
Li, Huawei [1 ]
Li, Xiaowei [1 ]
机构
[1] Chinese Acad Sci, Inst Comp Technol, State Key Lab Comp Architecture, Beijing 100190, Peoples R China
基金
中国国家自然科学基金;
关键词
Deep learning; error tolerance; neural network (NN); timing variation;
D O I
10.1109/TVLSI.2017.2682885
中图分类号
TP3 [计算技术、计算机技术];
学科分类号
0812 ;
摘要
Unlike conventional ICs, approximate computing chips are less sensitive to hardware errors. This fascinating feature can be utilized to improve the performance of chip design and even change the timing closure procedure of digital circuit design flow. In this paper, we study the potential of resilience-aware circuit clocking scheme, and demonstrate the methodology with advanced neural network (NN)-based accelerator. We propose a novel timing analysis and frequency setting method for NN-based approximate computing circuits based on in-field NN retraining. With the proposed iterative retiming-and-retraining framework, NN-based accelerator can be retrained to operate safely at aggressive operating frequencies compared with the frequency decided purely by statistical timing analysis or Monto Carlo analysis. For nanometer process technology with increasing threats of timing errors induced by process variation, noises, and so on, our retiming-and-retraining method enables higher circuit operating frequency and enables dynamic precision/frequency adjustment for approximate computing circuits. We evaluate the methodology with both the neural and deep learning accelerators in experiments. The experimental results show that timing errors in neural circuits can be effectively tamed for different applications, so that the circuits can operate at higher clocking rates under the specified quality constraint or be dynamically scaled to work at a wide range of frequency states with only minor accuracy losses.
引用
收藏
页码:2736 / 2748
页数:13
相关论文
共 50 条
  • [21] Neural Network-Based Frequency Optimization for Superconducting Quantum Chips
    BinHan Lu
    QingSong Li
    Peng Wang
    ZhaoYun Chen
    YuChun Wu
    GuoPing Guo
    Chinese Physics Letters, 2025, 42 (03) : 20 - 26
  • [22] Neural Network-Based Frequency Optimization for Superconducting Quantum Chips
    Lu, Bin-Han
    Li, Qing-Song
    Wang, Peng
    Chen, Zhao-Yun
    Wu, Yu-Chun
    Guo, Guo-Ping
    CHINESE PHYSICS LETTERS, 2025, 42 (03)
  • [23] Pole placement self-tuning feedforward control of PMSM with neural-network-based observer
    Li, Hong-Ru
    Wang, Jian-Hui
    Wang, Jue
    Gu, Shu-Sheng
    Kongzhi yu Juece/Control and Decision, 2001, 16 (SUPPL.): : 681 - 684
  • [24] Neural-network-based multistage interconnection network routing
    Krishnamoorthy, V
    Yi, P
    Zhang, YQ
    INTELLIGENT COMPUTING: THEORY AND APPLICATIONS, 2003, 5103 : 162 - 165
  • [25] Neural-network-based inverse hysteresis model
    Ma, Lian-Wei
    Tan, Yong-Hong
    Zou, Tao
    Kongzhi Lilun Yu Yingyong/Control Theory and Applications, 2008, 25 (05): : 823 - 826
  • [26] A compact neural-network-based CDMA receiver
    Chen, DC
    Sheu, BJ
    IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS II-ANALOG AND DIGITAL SIGNAL PROCESSING, 1998, 45 (03): : 384 - 387
  • [27] Approach to the neural-network-based data mining
    Zheng, Zhijun
    Lin, Xiaguang
    Zheng, Shouqi
    Xi'an Jianzhu Keji Daxue Xuebao/Journal of Xi'an University of Architecture & Technology, 2000, 32 (01): : 28 - 30
  • [28] A NEURAL-NETWORK-BASED DEDICATED THINNING METHOD
    AHMED, P
    PATTERN RECOGNITION LETTERS, 1995, 16 (06) : 585 - 590
  • [29] On the research of neural-network-based dynamics for robot
    Yang, Z
    Meng, ZD
    ADVANCES IN DYNAMICS, INSTRUMENTATION AND CONTROL, 2004, : 354 - 360
  • [30] A Neural-Network-Based Gaussian Nonlinear Filter
    Giraldo-Grueso, Felipe
    Popov, Andrey A.
    Zanetti, Renato
    AIAA SCITECH 2024 FORUM, 2024,