Improving energy consumption of pattern recognition by combining processor-centric and bio-inspired considerations

Yathindu R. Hettiarachchige, Asad I. Khan, Jan Carlo Barca

    Research output: Contribution to journalArticleResearchpeer-review

    Abstract

    This paper investigates aspects of bio-inspired models that help create more energy efficient methods in pattern recognition. A comparison between a biologically plausible pattern recognition approach and a purely computer based (algorithmic) approach yielded three main findings. Firstly, the occurrence of low-complexity parallel sub-processes within the bio-inspired approach allows higher energy efficiency by relaxing the requirement of having faster processors. Secondly, the bio-inspired approach takes advantage of numerous computationally inexpensive sub-processes that will scale better in massively parallel environments, such as neuromorphic computers, thus providing comparable speed. Finally, it is far more easier to adapt across a range of application domains than its algorithmic counterpart.

    Original languageEnglish
    Pages (from-to)54-63
    Number of pages10
    JournalBiologically Inspired Cognitive Architectures
    Volume23
    DOIs
    Publication statusPublished - 1 Jan 2018

    Keywords

    • Bio-inspired
    • Distributed pattern recognition
    • Parallelism
    • Strong AI

    Cite this