User Tools

Site Tools


projects:year10:10a.008.ul

10a.008.UL - Decentralized and Distributed Deep Learning for Industrial IoT Devices

Project - Summary

GOALS

  • Optimize the energy cost occurred by the communication and enhance performance efficiency of deploying deep learning on Industrial IoT devices.
  • Identify application-specific characteristics of deep learning applications and leverage algorithm-system codesign to offer cross-layer solutions that make intelligence faster and more cost-efficient.

Project - Team

Team Member Role Email Phone Number Academic Site/IAB
Xiali (Sharon) Hei PI xiali.hei@louisiana.edu (337) 482-1037 UL Lafayette
Li Chen Co-PI li.chen@louisiana.edu N/A UL Lafayette
Simin Javaherian Student N/A N/A UL Lafayette
Sai Venkatesh Chilukoti Student N/A N/A UL Lafayette
Shovon Paul Student N/A N/A UL Lafayette

Project - Novelty of Approach

  • Previous schemes would lower the accuracy, and some of them focus on memory reduction, we will work on reducing the number of the parameters to be updated and parameter updating round while maintaining the desired accuracy.
  • Application-specific characteristics have not been fully exploited in the design of an end-to-end deep learning pipeline. We will incorporate them in the algorithm-system codesign, offering novel cross-layer techniques.

Project - Deliverables

Deliverables
1 Memory/Computing cost evaluation
2 Energy/Communication cost evaluation
3 Design of system architecture
4 Design of techniques and unit test

Project - Benefits to IAB

  • If successful, such as understanding of predictive IoT would further allow low computing deep learning on small IoT devices to be accomplished, which would have a high impact on the field of medical devices, energy, and industrial robotics.
  • This proposed innovative solution as a new platform will benefit the broad AI society, pushing the technology boundary, and enabling a better understanding of deep learning applications and their underlying systems.
  • These new studies may provide insights that could lead to novel computing approaches and develop methods that can be applied to IoT devices and other related techniques.

Project - Documents

projects/year10/10a.008.ul.txt · Last modified: 2022/05/10 15:58 by sally.johnson