Alternate pulse scheme in a hardware neural network for reducing the influence of asymmetry on synaptic weight updating

Cheng Li, Junseop An, Jun Young Kweon, Yun Heub Song

Research output: Contribution to journalArticle

Abstract

Recently, researchers have been paying more attention to hardware neural networks with memory devices acting as synapses due to their small area and low power consumption. We prefer to use a phase-change memory (PCM) device as a synapse due to its excellent synapse-like properties. However, due to the programming mechanism of PCMs, the asymmetric characteristic of the PCM between partially set and partially reset is severe. In this paper, we proposed an alternate pulse scheme (APS) to reduce the influence of the asymmetry between long-term potentiation and long-term depression operations.

Original languageEnglish
Pages (from-to)119-126
Number of pages8
JournalJournal of Semiconductor Technology and Science
Volume20
Issue number1
DOIs
StatePublished - 2020 Feb

    Fingerprint

Keywords

  • Feature extraction
  • Neural network
  • Neuron
  • PCM
  • Phase-change memory
  • Synapse

Cite this