On-device deep multi-task inference via multi-task zipping
Future mobile devices are anticipated to perceive, understand and react to the world on their own by running multiple correlated deep neural networks locally on-device. Yet the complexity of these deep models needs to be trimmed down both within-model and cross-model to fit in mobile storage and mem...
Saved in:
Main Authors: | , , , , , |
---|---|
Format: | text |
Language: | English |
Published: |
Institutional Knowledge at Singapore Management University
2023
|
Subjects: | |
Online Access: | https://ink.library.smu.edu.sg/sis_research/6724 https://ink.library.smu.edu.sg/context/sis_research/article/7727/viewcontent/tmc21_he.pdf |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Institution: | Singapore Management University |
Language: | English |
id |
sg-smu-ink.sis_research-7727 |
---|---|
record_format |
dspace |
spelling |
sg-smu-ink.sis_research-77272024-03-06T03:26:57Z On-device deep multi-task inference via multi-task zipping HE, Xiaoxi WANG, Xu ZHOU, Zimu WU, Jiahang YANG, Zheng THIELE, Lothar Future mobile devices are anticipated to perceive, understand and react to the world on their own by running multiple correlated deep neural networks locally on-device. Yet the complexity of these deep models needs to be trimmed down both within-model and cross-model to fit in mobile storage and memory. Previous studies squeeze the redundancy within a single model. In this work, we aim to reduce the redundancy across multiple models. We propose Multi-Task Zipping (MTZ), a framework to automatically merge correlated, pre-trained deep neural networks for cross-model compression. Central in MTZ is a layer-wise neuron sharing and incoming weight updating scheme that induces a minimal change in the error function. MTZ inherits information from each model and demands light retraining to re-boost the accuracy of individual tasks. MTZ supports typical network layers (fully-connected, convolutional and residual) and applies to inference tasks with different input domains. Evaluations show that MTZ can fully merge the hidden layers of two VGG-16 networks with a 3.18% increase in the test error averaged on ImageNet for object classification and CelebA for facial attribute classification, or share 39.61% parameters between the two networks with 2023-05-01T07:00:00Z text application/pdf https://ink.library.smu.edu.sg/sis_research/6724 info:doi/10.1109/TMC.2021.3124306 https://ink.library.smu.edu.sg/context/sis_research/article/7727/viewcontent/tmc21_he.pdf http://creativecommons.org/licenses/by-nc-nd/4.0/ Research Collection School Of Computing and Information Systems eng Institutional Knowledge at Singapore Management University Deep Neural Networks Model Compression Multi-Task Learning OS and Networks Software Engineering |
institution |
Singapore Management University |
building |
SMU Libraries |
continent |
Asia |
country |
Singapore Singapore |
content_provider |
SMU Libraries |
collection |
InK@SMU |
language |
English |
topic |
Deep Neural Networks Model Compression Multi-Task Learning OS and Networks Software Engineering |
spellingShingle |
Deep Neural Networks Model Compression Multi-Task Learning OS and Networks Software Engineering HE, Xiaoxi WANG, Xu ZHOU, Zimu WU, Jiahang YANG, Zheng THIELE, Lothar On-device deep multi-task inference via multi-task zipping |
description |
Future mobile devices are anticipated to perceive, understand and react to the world on their own by running multiple correlated deep neural networks locally on-device. Yet the complexity of these deep models needs to be trimmed down both within-model and cross-model to fit in mobile storage and memory. Previous studies squeeze the redundancy within a single model. In this work, we aim to reduce the redundancy across multiple models. We propose Multi-Task Zipping (MTZ), a framework to automatically merge correlated, pre-trained deep neural networks for cross-model compression. Central in MTZ is a layer-wise neuron sharing and incoming weight updating scheme that induces a minimal change in the error function. MTZ inherits information from each model and demands light retraining to re-boost the accuracy of individual tasks. MTZ supports typical network layers (fully-connected, convolutional and residual) and applies to inference tasks with different input domains. Evaluations show that MTZ can fully merge the hidden layers of two VGG-16 networks with a 3.18% increase in the test error averaged on ImageNet for object classification and CelebA for facial attribute classification, or share 39.61% parameters between the two networks with |
format |
text |
author |
HE, Xiaoxi WANG, Xu ZHOU, Zimu WU, Jiahang YANG, Zheng THIELE, Lothar |
author_facet |
HE, Xiaoxi WANG, Xu ZHOU, Zimu WU, Jiahang YANG, Zheng THIELE, Lothar |
author_sort |
HE, Xiaoxi |
title |
On-device deep multi-task inference via multi-task zipping |
title_short |
On-device deep multi-task inference via multi-task zipping |
title_full |
On-device deep multi-task inference via multi-task zipping |
title_fullStr |
On-device deep multi-task inference via multi-task zipping |
title_full_unstemmed |
On-device deep multi-task inference via multi-task zipping |
title_sort |
on-device deep multi-task inference via multi-task zipping |
publisher |
Institutional Knowledge at Singapore Management University |
publishDate |
2023 |
url |
https://ink.library.smu.edu.sg/sis_research/6724 https://ink.library.smu.edu.sg/context/sis_research/article/7727/viewcontent/tmc21_he.pdf |
_version_ |
1794549748643921920 |