FEIT Research Project Database

Coded computation for machine learning

Project Leader: Jingge Zhu
Primary Contact: Jingge Zhu (jingge.zhu@unimelb.edu.au)
Keywords: distributed computing; machine learning
Disciplines: Computing and Information Systems,Electrical & Electronic Engineering

Coded computation is an emerging technique which uses error correction codes to improve the reliability of distributed computation systems. Roughly speaking, a big computational task is divided into smaller subtasks along with additional redundancies. Coded computation schemes guarantee that the overall computational tasks can be accomplished even if some of the smaller subtasks are not completed. In this project, we will study the coded computation of general functions using deep neural networks and its application for solving large-scale machine learning problems.