Compression of Activation Signals from Split Deep Neural Network

Flavio Brito, Lucas Silva, Leonardo Ramalho, Silvia Lins, Neiva Linder, Aldebaro Klautau

Research output: Chapter in Book/Report/Conference proceedingPaper in conference proceedingpeer-review

Abstract

The use of artificial neural networks for the purpose of image classification, together with the advancement in computational capabilities of edge devices, plays an important role in the new emerging 5G use case scenarios. However, one of the main challenges of applications involving the use of these networks in edge devices is still the limitation of computational resources. An alternative for saving resources and promoting privacy are the split learning (or split inference) techniques, in which a deep neural network is cut into two parts and executed in distinct devices. Most of these techniques rely on sending the activation signals (output of the cut layer) through the communication channel. This work proposes a new compression algorithm for decreasing the bit rate required for the transmission of the activation signals (or simply 'scores'). The presented results demonstrate that the transmission rate can be decreased without hurting the neural network accuracy.

Original languageEnglish
Title of host publication2022 IEEE Latin-American Conference on Communications, LATINCOM 2022
EditorsIgor M. Moraes, Miguel Elias M. Campista, Yacine Ghamri-Doudane, Costa Luis Henrique M. K. Costa, Marcelo G. Rubinstein
PublisherIEEE - Institute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9781665482257
DOIs
Publication statusPublished - 2022
Event14th IEEE Latin-American Conference on Communications, LATINCOM 2022 - Rio de Janeiro, Brazil
Duration: 2022 Nov 302022 Dec 2

Publication series

Name2022 IEEE Latin-American Conference on Communications, LATINCOM 2022

Conference

Conference14th IEEE Latin-American Conference on Communications, LATINCOM 2022
Country/TerritoryBrazil
CityRio de Janeiro
Period2022/11/302022/12/02

Subject classification (UKÄ)

  • Computer Engineering

Free keywords

  • deep neural networks
  • image classification
  • scores compression
  • Split learning

Fingerprint

Dive into the research topics of 'Compression of Activation Signals from Split Deep Neural Network'. Together they form a unique fingerprint.

Cite this