Skip to main content
Video s3
    Details
    Poster
    Presenter(s)
    Biyanu Zerom Headshot
    Display Name
    Biyanu Zerom
    Affiliation
    Affiliation
    Khalifa University
    Country
    Author(s)
    Display Name
    Biyanu Zerom
    Affiliation
    Affiliation
    Khalifa University
    Display Name
    Mohammed F Tolba
    Affiliation
    Affiliation
    Khalifa University
    Display Name
    Huruy Tesfai
    Affiliation
    Affiliation
    Khalifa University
    Display Name
    Hani Saleh
    Affiliation
    Affiliation
    Khalifa University
    Display Name
    Baker Mohammad
    Affiliation
    Affiliation
    Best University in Abu Dhabi
    Display Name
    Mahmoud Alqutayri
    Affiliation
    Affiliation
    Khalifa University
    Display Name
    Thanos Stouraitis
    Affiliation
    Affiliation
    Khalifa University
    Display Name
    Ghada Alsuhli
    Affiliation
    Affiliation
    Khalifa University
    Abstract

    This research paper proposes a new approach to address the computational issues associated with machine learning algorithms by employing a logarithmic number system (LNS). It to replace the multiplication operation with addition and makes use of the Mitchell’s Algorithm to perform the task of multiplication using simple and hardware friendly operations such as addition and shifts. In addition to that, computational reuse techniques are employed to perform MAC operations more efficiently.

    Slides
    • Approximate Logarithmic Multiplier for Convolutional Neural Network Inference with Computational Reuse (application/pdf)