Hardware-compatible and energy-efficient neural network for solving machine learning tasks

Details

Supervisors

Abreu Araujo, Flavio

Faculty

Ecole polytechnique de Louvain

Degree label

Master [120] : ingénieur civil électricien, à finalité spécialisée

Abstract

This study explores the fusion of machine learning, particularly reservoir computing networks (RCNs), with spintronics to achieve energy-efficient hardware implementations. The focus is on utilizing Spin Torque Vortex Oscillators (STVOs) as “neurons” within a reservoir computing network. By leveraging the unique properties of STVOs, the research aims to assess their performance in comparison to other hardware-implemented and conventional reservoir computing networks. The investigation includes a comprehensive analysis of STVO-specific parameters and their impact on the overall system’s efficiency and computational capability. This work contributes to the advancement of energy-efficient machine learning hardware, demonstrating the potential of STVOs in enhancing the performance of reservoir computing networks.