DLR-Logo -> http://www.dlr.de
DLR Portal Home | Imprint | Privacy Policy | Contact | Deutsch
Fontsize: [-] Text [+]

Using Network Architecture Search for Optimizing Tensor Compression

Thirunavukkarasu, Arunachalam and Helms, Domenik (2023) Using Network Architecture Search for Optimizing Tensor Compression. In: Designing Modern Embedded Systems: Software, Hardware, and Applications. Springer. International Embedded Systems Symposium, Lippstadt. doi: 10.1007/978-3-031-34214-1_12.

[img] PDF - Only accessible within DLR


In this work we propose to use Network Architecture Search (NAS) for controlling the per layer parameters of a Tensor Compression (TC) algorithm using Tucker decomposition in order to optimize a given convolutional neural network for its parameter count and thus inference performance on embedded systems. TC enables a quick generation of the next instance in the NAS process, avoiding the need for a time consuming full training after each step. We show that this approach is more eficient than conventional NAS and can outperform all TC heuristics reported so far. Nevertheless it is still a very time consuming process, finding a good solution in the vast search space of layer-wise TC. We show that, it is possible to reduce the parameter size upto 85% for the cost of 0.1- 1% of Top-1 accuracy on our vision processing benchmarks. Further, it is shown that the compressed model occupies just 20% of the original memory size which is required for storing the entire uncompressed model, with an increase in the inference speed of upto 2.5 times without much loss in the performance indicating potential gains for embedded systems.

Item URL in elib:https://elib.dlr.de/196697/
Document Type:Conference or Workshop Item (Speech)
Title:Using Network Architecture Search for Optimizing Tensor Compression
AuthorsInstitution or Email of AuthorsAuthor's ORCID iDORCID Put Code
Thirunavukkarasu, ArunachalamUNSPECIFIEDhttps://orcid.org/0000-0003-0824-140XUNSPECIFIED
Helms, DomenikUNSPECIFIEDhttps://orcid.org/0000-0001-7326-200XUNSPECIFIED
Date:11 June 2023
Journal or Publication Title:Designing Modern Embedded Systems: Software, Hardware, and Applications
Refereed publication:Yes
Open Access:No
Gold Open Access:No
In ISI Web of Science:No
EditorsEmailEditor's ORCID iDORCID Put Code
Series Name:IESS
Keywords:Tensor Compression · Embedded systems · Network Architecture Search · Tucker Decomposition · Convolutional Neural Network.
Event Title:International Embedded Systems Symposium
Event Location:Lippstadt
Event Type:international Conference
HGF - Research field:Aeronautics, Space and Transport
HGF - Program:Transport
HGF - Program Themes:Road Transport
DLR - Research area:Transport
DLR - Program:V ST Straßenverkehr
DLR - Research theme (Project):V - V&V4NGC - Methoden, Prozesse und Werkzeugketten für die Validierung & Verifikation von NGC
Location: Oldenburg
Institutes and Institutions:Institute of Systems Engineering for Future Mobility
Deposited By: Helms, Domenik
Deposited On:31 Aug 2023 07:42
Last Modified:31 Aug 2023 07:42

Repository Staff Only: item control page

Help & Contact
electronic library is running on EPrints 3.3.12
Website and database design: Copyright © German Aerospace Center (DLR). All rights reserved.