Structured Pruning in Deep Neural Networks with Trainable Probability Masks

F. Martinini, A. Enttsel, A. Marchioni, M. Mangia, R. Rovatti, G. Setti

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

The current trend of over-parameterized Deep Neural Networks makes the deployment on resource constrained systems challenging. To deal with this, optimization techniques, such as network pruning, can be adopted. We propose a novel pruning technique based on trainable probability masks that, when binarized, select the elements of the network to prune. Our method features i) an automatic selections of the elements to prune by jointly training the binary masks with the model, ii) the capability of controlling the pruning level through hyper-parameters of a novel regularization term. We assess the effectiveness of our method by employing it in the structured pruning of the fully connected layers of shallow and deep neural networks where it outperforms the magnitude-based pruning approaches.

Original languageEnglish (US)
Title of host publication2023 IEEE 66th International Midwest Symposium on Circuits and Systems, MWSCAS 2023
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages1020-1024
Number of pages5
ISBN (Electronic)9798350302103
DOIs
StatePublished - 2023
Event2023 IEEE 66th International Midwest Symposium on Circuits and Systems, MWSCAS 2023 - Tempe, United States
Duration: Aug 6 2023Aug 9 2023

Publication series

NameMidwest Symposium on Circuits and Systems
ISSN (Print)1548-3746

Conference

Conference2023 IEEE 66th International Midwest Symposium on Circuits and Systems, MWSCAS 2023
Country/TerritoryUnited States
CityTempe
Period08/6/2308/9/23

Keywords

  • Deep Neural Network
  • Network Pruning
  • Probability Mask
  • Trainable Binary Mask

ASJC Scopus subject areas

  • Electronic, Optical and Magnetic Materials
  • Electrical and Electronic Engineering

Fingerprint

Dive into the research topics of 'Structured Pruning in Deep Neural Networks with Trainable Probability Masks'. Together they form a unique fingerprint.

Cite this