Low Cost and Fault Tolerant Parallel Computing Using Stochastic Two-Dimensional Finite State Machine

Xuechun WANG; Yuan JI; Wendong CHEN; Feng RAN; Aiying GUO

doi:10.1587/transinf.2017PAL0003

Low Cost and Fault Tolerant Parallel Computing Using Stochastic Two-Dimensional Finite State Machine

Xuechun WANG, Yuan JI, Wendong CHEN, Feng RAN, Aiying GUO

Full Text Views

0

Share
Cite this

Summary :

Hardware implementation of neural networks usually have high computational complexity that increase exponentially with the size of a circuit, leading to more uncertain and unreliable circuit performance. This letter presents a novel Radial Basis Function (RBF) neural network based on parallel fault tolerant stochastic computing, in which number is converted from deterministic domain to probabilistic domain. The Gaussian RBF for middle layer neuron is implemented using stochastic structure that reduce the hardware resources significantly. Our experimental results from two pattern recognition tests (the Thomas gestures and the MIT faces) show that the stochastic design is capable to maintain equivalent performance when the stream length set to 10Kbits. The stochastic hidden neuron uses only 1.2% hardware resource compared with the CORDIC algorithm. Furthermore, the proposed algorithm is very flexible in design tradeoff between computing accuracy, power consumption and chip area.

Publication: IEICE TRANSACTIONS on Information Vol.E100-D No.12 pp.2866-2870

Publication Date: 2017/12/01

Publicized: 2017/07/18

Online ISSN: 1745-1361

DOI: 10.1587/transinf.2017PAL0003

Type of Manuscript: Special Section LETTER (Special Section on Parallel and Distributed Computing and Networking)

Category: Architecture

Authors

Xuechun WANG
  Shanghai University
Yuan JI
  Shanghai University
Wendong CHEN
  Shanghai University
Feng RAN
  Shanghai University
Aiying GUO
  Shanghai University

Keyword

stochastic computing, fault tolerant, pattern recognition, power consumption

Cite this

Copy

Xuechun WANG, Yuan JI, Wendong CHEN, Feng RAN, Aiying GUO, "Low Cost and Fault Tolerant Parallel Computing Using Stochastic Two-Dimensional Finite State Machine" in IEICE TRANSACTIONS on Information, vol. E100-D, no. 12, pp. 2866-2870, December 2017, doi: 10.1587/transinf.2017PAL0003.
Abstract: Hardware implementation of neural networks usually have high computational complexity that increase exponentially with the size of a circuit, leading to more uncertain and unreliable circuit performance. This letter presents a novel Radial Basis Function (RBF) neural network based on parallel fault tolerant stochastic computing, in which number is converted from deterministic domain to probabilistic domain. The Gaussian RBF for middle layer neuron is implemented using stochastic structure that reduce the hardware resources significantly. Our experimental results from two pattern recognition tests (the Thomas gestures and the MIT faces) show that the stochastic design is capable to maintain equivalent performance when the stream length set to 10Kbits. The stochastic hidden neuron uses only 1.2% hardware resource compared with the CORDIC algorithm. Furthermore, the proposed algorithm is very flexible in design tradeoff between computing accuracy, power consumption and chip area.
URL: https://globals.ieice.org/en_transactions/information/10.1587/transinf.2017PAL0003/_p

Copy

@ARTICLE{e100-d_12_2866,
author={Xuechun WANG, Yuan JI, Wendong CHEN, Feng RAN, Aiying GUO, },
journal={IEICE TRANSACTIONS on Information},
title={Low Cost and Fault Tolerant Parallel Computing Using Stochastic Two-Dimensional Finite State Machine},
year={2017},
volume={E100-D},
number={12},
pages={2866-2870},
abstract={Hardware implementation of neural networks usually have high computational complexity that increase exponentially with the size of a circuit, leading to more uncertain and unreliable circuit performance. This letter presents a novel Radial Basis Function (RBF) neural network based on parallel fault tolerant stochastic computing, in which number is converted from deterministic domain to probabilistic domain. The Gaussian RBF for middle layer neuron is implemented using stochastic structure that reduce the hardware resources significantly. Our experimental results from two pattern recognition tests (the Thomas gestures and the MIT faces) show that the stochastic design is capable to maintain equivalent performance when the stream length set to 10Kbits. The stochastic hidden neuron uses only 1.2% hardware resource compared with the CORDIC algorithm. Furthermore, the proposed algorithm is very flexible in design tradeoff between computing accuracy, power consumption and chip area.},
keywords={},
doi={10.1587/transinf.2017PAL0003},
ISSN={1745-1361},
month={December},}

Copy

TY - JOUR
TI - Low Cost and Fault Tolerant Parallel Computing Using Stochastic Two-Dimensional Finite State Machine
T2 - IEICE TRANSACTIONS on Information
SP - 2866
EP - 2870
AU - Xuechun WANG
AU - Yuan JI
AU - Wendong CHEN
AU - Feng RAN
AU - Aiying GUO
PY - 2017
DO - 10.1587/transinf.2017PAL0003
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E100-D
IS - 12
JA - IEICE TRANSACTIONS on Information
Y1 - December 2017
AB - Hardware implementation of neural networks usually have high computational complexity that increase exponentially with the size of a circuit, leading to more uncertain and unreliable circuit performance. This letter presents a novel Radial Basis Function (RBF) neural network based on parallel fault tolerant stochastic computing, in which number is converted from deterministic domain to probabilistic domain. The Gaussian RBF for middle layer neuron is implemented using stochastic structure that reduce the hardware resources significantly. Our experimental results from two pattern recognition tests (the Thomas gestures and the MIT faces) show that the stochastic design is capable to maintain equivalent performance when the stream length set to 10Kbits. The stochastic hidden neuron uses only 1.2% hardware resource compared with the CORDIC algorithm. Furthermore, the proposed algorithm is very flexible in design tradeoff between computing accuracy, power consumption and chip area.
ER -