Presentation Details
  1. Home
  2. Presentation List
  3. Presentation Details
Fault Modeling and Testing of Spiking Neural Network Chips
ID：48 View Protection：PUBLIC Updated Time：2021-08-19 20:37:15 Hits：656 Oral Presentation

Start Time：2021-08-19 20:40（Asia/Shanghai）

Duration：20min

Session：RS Regular Paper Session » RS1A1. When Machine Learning Meets Testing and Security

Video Play Presentation File Attachment File
Abstract
Spiking neural network (SNN) is a very promising low-power neural network that can be implemented in asynchronous circuits.  However, it is hard to test SNN chips since they are inherently probabilistic and fault tolerant.  So far, there is no good fault model and test methodology suitable for SNN chips.  In this paper, we propose seven behavior fault models for SNN based on the function of neurons and synapses.  We also propose a test methodology, which considers the output response as a distribution rather than specific values.  The experiment results on a MNIST dataset show that although SNN is fault tolerant, two fault models are still critical for SNN chips.  Given the digit recognition application, the accuracy of chips that passed our test is 88.90%, which is indistinguishable from that of good chips, even in the effects of random seeds.
 
Keywords
Spiking neural network;Asynchronous circuits;Fault model;Fault simulation;test methodology
Speaker
I-Wei Chiu
National Taiwan University

I-Wei Chiu received his BSEE degree from National Taiwan Normal University, Taipei, Taiwan, in 2020, where he currently is working toward the MSEE degree in electrical engineering. His research interests include neuromorphic circuits testing.

Li James Chien Mo
National Taiwan University

James Chien-Mo Li received his BSEE degree in 1993 from National Taiwan University, Taipei, Taiwan.  He received his MSEE and PhD degrees in electrical engineering from Stanford University in 1997 and 2002 respectively. He is currently a professor of Graduate Institute of Electronics Engineering, National Taiwan University, Taipei, Taiwan.  His research interest includes test generation, low power testing, flexible electronics, and diagnosis. He is a member of the IEEE.

Submit Comment
Verify Code Change Another
All Comments
Important Date

  • Conference Date

    Aug 18

    2021

    to

    Aug 20

    2021

  • May 10 2021

    Draft paper submission deadline

  • Aug 16 2021

    Early Bird Registration

  • Aug 19 2021

    Contribution Submission Deadline

  • Aug 20 2021

    Registration deadline

Sponsored By
IEEE
Tongji University
Chinese Computer Federation
Organized By
Tongji University
Contact Information
Login to view full contact details
Previous Conferences