Efficient and robust bitstream processing in binarised neural networks

Sercan Aygun*, Ece Olcay Gunes, Christophe De Vleeschouwer

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

4 Citations (Scopus)

Abstract

In the neural network context, used in a variety of applications, binarised networks, which describe both weights and activations as single-bit binary values, provide computationally attractive solutions. A lightweight binarised neural network system can be constructed using only logic gates and counters together with a two-valued activation function unit. However, binarised neural networks represent the weights and the neuron outputs with only one bit, making them sensitive to bit-flipping errors. Binarised weights and neurons are manipulated by the utilisation of bitstream processing with regard to stochastic computing to cope with this error sensitivity. Stochastic computing is shown to provide robustness for bit errors on data while being built on a hardware structure, whose implementation is simplified by a novel subtraction-free implementation of the neuron activation.

Original languageEnglish
Pages (from-to)219-222
Number of pages4
JournalElectronics Letters
Volume57
Issue number5
DOIs
Publication statusPublished - Mar 2021

Bibliographical note

Publisher Copyright:
© 2021 The Authors. Electronics Letters published by John Wiley & Sons Ltd on behalf of The Institution of Engineering and Technology.

Keywords

  • Logic circuits
  • Logic elements
  • Neural net devices
  • Neural nets (circuit implementations)

Fingerprint

Dive into the research topics of 'Efficient and robust bitstream processing in binarised neural networks'. Together they form a unique fingerprint.

Cite this