MemSE: Fast MSE Prediction for Noisy Memristor-Based DNN Accelerators

被引:0
|
作者
Kern, Jonathan [1 ,2 ]
Henwood, Sebastien [1 ]
Mordido, Goncalo [1 ,3 ]
Dupraz, Elsa [2 ]
Aissa-El-Bey, Abdeldjalil [2 ]
Savaria, Yvon [1 ]
Leduc-Primeau, Francois [1 ]
机构
[1] Polytech Montreal, Dept Elect Engn, Montreal, PQ, Canada
[2] IMT Atlantique, Lab STICC, CNRS UMR 6285, Brest, France
[3] Mila Quebec AI Inst, Montreal, PQ, Canada
关键词
OPTIMIZATION;
D O I
10.1109/AICAS54282.2022.9869978
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Memristors enable the computation of matrix-vector multiplications (MVM) in memory and, therefore, show great potential in highly increasing the energy efficiency of deep neural network (DNN) inference accelerators. However, computations in memristors suffer from hardware non-idealities and are subject to different sources of noise that may negatively impact system performance. In this work, we theoretically analyze the mean squared error of DNNs that use memristor crossbars to compute MVM. We take into account both the quantization noise, due to the necessity of reducing the DNN model size, and the programming noise, stemming from the variability during the programming of the memristance value. Simulations on pre-trained DNN models showcase the accuracy of the analytical prediction. Furthermore the proposed method is almost two order of magnitude faster than Monte-Carlo simulation, thus making it possible to optimize the implementation parameters to achieve minimal error for a given power constraint.
引用
收藏
页码:62 / 65
页数:4
相关论文
共 28 条
  • [21] Self-organizations with fast-slow time scale in a memristor-based Shinriki's circuit
    Rao, Xiao-Bo
    Zhao, Xu-Ping
    Gao, Jian-She
    Zhang, Jian-Gang
    COMMUNICATIONS IN NONLINEAR SCIENCE AND NUMERICAL SIMULATION, 2021, 94
  • [22] Memristor-based circuit design of episodic memory neural network and its application in hurricane category prediction
    Wan, Qiuzhen
    Liu, Jiong
    Liu, Tieqiao
    Sun, Kunliang
    Qin, Peng
    NEURAL NETWORKS, 2024, 174
  • [23] MSE Prediction in DCT-based Lossy Compression of Noise-Free and Noisy Remote Sensing
    Krivenko, Serge
    Zriakhov, Mikhail
    Lukin, Vladimir
    Vozel, Benoit
    2018 14TH INTERNATIONAL CONFERENCE ON ADVANCED TRENDS IN RADIOELECTRONICS, TELECOMMUNICATIONS AND COMPUTER ENGINEERING (TCSET), 2018, : 883 - 888
  • [24] Fast EIS acquisition method based on SSA-DNN prediction model
    Chang, Chun
    Pan, Yaliang
    Wang, Shaojin
    Jiang, Jiuchun
    Tian, Aina
    Gao, Yang
    Jiang, Yan
    Wu, Tiezhou
    ENERGY, 2024, 288
  • [25] Enhanced memristor-based gas sensor for fast detection using a porous carbon nanotube top electrode with membrane
    Chae, Myoungsu
    Lee, Doowon
    Jung, Jinsu
    Kim, Hee-Dong
    CELL REPORTS PHYSICAL SCIENCE, 2023, 4 (11):
  • [26] LOMA: Fast Auto-Scheduling on DNN Accelerators through Loop-Order-based Memory Allocation
    Symons, Arne
    Mei, Linyan
    Verhelst, Marian
    2021 IEEE 3RD INTERNATIONAL CONFERENCE ON ARTIFICIAL INTELLIGENCE CIRCUITS AND SYSTEMS (AICAS), 2021,
  • [27] Fast fixed-time sliding mode control for synchronization of chaotic systems with unmodeled dynamics and disturbance; applied to memristor-based oscillator
    Mirzaei, Mohammad Javad
    Aslmostafa, Ehsan
    Asadollahi, Mostafa
    Padar, Naser
    JOURNAL OF VIBRATION AND CONTROL, 2023, 29 (9-10) : 2129 - 2143
  • [28] Fast Reservoir Characterization with AI-Based Lithology Prediction Using Drill Cuttings Images and Noisy Labels
    Tolstaya, Ekaterina
    Shakirov, Anuar
    Mezghani, Mokhles
    Safonov, Sergey
    JOURNAL OF IMAGING, 2023, 9 (07)