Impact of Disentanglement on Pruning Neural Networks

descriptionPublicationkeyboard_double_arrow_right Conference object , Other literature type , Article , Preprint 01 Jan 2023Embargo end date: 01 Jan 2023Publisher:ZenodoJournal:CoRR, volume abs/2307.09994

Authors: Carl Shneider; Peyman Rostami; Anis Kacem; Nilotpal Sinha; Abdelrahman Shabayek; Djamila Aouada;

doi: 10.5281/zenodo.8060029 , 10.5281/zenodo.8060030 , 10.48550/arxiv.2307.09994

arXiv: 2307.09994

Impact of Disentanglement on Pruning Neural Networks

- Summary
- Subjects
- Metrics

Abstract

Efficient model compression techniques are required to deploy deep neural networks (DNNs) on edge devices for task specific objectives. A variational autoencoder (VAE) framework is combined with a pruning criterion to investigate the impact of having the network learn disentangled representations on the pruning process for the classification task. Poster from the Computer Vision, Imaging, and Machine Intelligence Research Group (CVI2) at SnT, University of Luxembourg. Selected for poster presentation during the first edition of the International Symposium on Computational Sensing ISCS23 in Luxembourg.

This work was funded by the Luxembourg National Research Fund (FNR) under the project reference C21/IS/15965298/ELITE.

Related Organizations

University of Luxembourg
Luxembourg

Keywords

Signal Processing (eess.SP), FOS: Computer and information sciences, Computer Science - Machine Learning, Computer Vision and Pattern Recognition (cs.CV), Neural Network Pruning, Computer Science - Computer Vision and Pattern Recognition, Deep learning, Edge computing, Machine Learning (cs.LG), Disentangled latent representation, Variational Autoencoder, FOS: Electrical engineering, electronic engineering, information engineering, Neural Network Compression, Electrical Engineering and Systems Science - Signal Processing

Impact byBIP!

	selected citations These citations are derived from selected sources. This is an alternative to the "Influence" indicator, which also reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).	0
	popularity This indicator reflects the "current" impact/attention (the "hype") of an article in the research community at large, based on the underlying citation network.	Average
	influence This indicator reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).	Average
	impulse This indicator reflects the initial momentum of an article directly after its publication, based on the underlying citation network.	Average