LifeLearner: Hardware-aware meta continual learning system for embedded computing platforms
LifeLearner: Hardware-aware meta continual learning system for embedded computing platforms
Continual Learning (CL) allows applications such as user personalization and household robots to learn on the fly and adapt to context. This is an important feature when context, actions, and users change. However, enabling CL on resource-constrained embedded systems is challenging due to the limited labeled data, memory, and computing capacity.In this paper, we propose LifeLearner, a hardware-aware meta continual learning system that drastically optimizes system resources (lower memory, latency, energy consumption) while ensuring high accuracy. Specifically, we (1) exploit meta-learning and rehearsal strategies to explicitly cope with data scarcity issues and ensure high accuracy, (2) effectively combine lossless and lossy compression to significantly reduce the resource requirements of CL and rehearsal samples, and (3) developed hardware-aware system on embedded and IoT platforms considering the hardware characteristics.As a result, LifeLearner achieves near-optimal CL performance, falling short by only 2.8% on accuracy compared to an Oracle baseline. With respect to the state-of-the-art (SOTA) Meta CL method, LifeLearner drastically reduces the memory footprint (by 178.7×), end-to-end latency by 80.8 - 94.2%, and energy consumption by 80.9 - 94.2%. In addition, we successfully deployed LifeLearner on two edge devices and a microcontroller unit, thereby enabling efficient CL on resource-constrained platforms where it would be impractical to run SOTA methods and the far-reaching deployment of adaptable CL in a ubiquitous manner. Code is available at https://github.com/theyoungkwon/LifeLearner.
continual learning, edge computing, latent replay, meta learning, microcontrollers, on-device training, product quantization
138-151
Association for Computing Machinery
Kwon, Young D.
3e8c3dcd-214c-4771-90f4-b36ede48d763
Chauhan, Jagmohan
831a12dc-6df9-40ea-8bb3-2c5da8882804
Jia, Hong
2acc4da7-3d5b-4d4b-a55c-a639b52b7942
Venieris, Stylianos I.
1dabc04a-7627-4a00-b1e7-80683682d51f
Mascolo, Cecilia
e4a7bcf7-72c8-43b7-b6b3-4f8980da245d
12 November 2023
Kwon, Young D.
3e8c3dcd-214c-4771-90f4-b36ede48d763
Chauhan, Jagmohan
831a12dc-6df9-40ea-8bb3-2c5da8882804
Jia, Hong
2acc4da7-3d5b-4d4b-a55c-a639b52b7942
Venieris, Stylianos I.
1dabc04a-7627-4a00-b1e7-80683682d51f
Mascolo, Cecilia
e4a7bcf7-72c8-43b7-b6b3-4f8980da245d
Kwon, Young D., Chauhan, Jagmohan, Jia, Hong, Venieris, Stylianos I. and Mascolo, Cecilia
(2023)
LifeLearner: Hardware-aware meta continual learning system for embedded computing platforms.
In SenSys 2023 - Proceedings of the 21st ACM Conference on Embedded Networked Sensors Systems.
Association for Computing Machinery.
.
(doi:10.1145/3625687.3625804).
Record type:
Conference or Workshop Item
(Paper)
Abstract
Continual Learning (CL) allows applications such as user personalization and household robots to learn on the fly and adapt to context. This is an important feature when context, actions, and users change. However, enabling CL on resource-constrained embedded systems is challenging due to the limited labeled data, memory, and computing capacity.In this paper, we propose LifeLearner, a hardware-aware meta continual learning system that drastically optimizes system resources (lower memory, latency, energy consumption) while ensuring high accuracy. Specifically, we (1) exploit meta-learning and rehearsal strategies to explicitly cope with data scarcity issues and ensure high accuracy, (2) effectively combine lossless and lossy compression to significantly reduce the resource requirements of CL and rehearsal samples, and (3) developed hardware-aware system on embedded and IoT platforms considering the hardware characteristics.As a result, LifeLearner achieves near-optimal CL performance, falling short by only 2.8% on accuracy compared to an Oracle baseline. With respect to the state-of-the-art (SOTA) Meta CL method, LifeLearner drastically reduces the memory footprint (by 178.7×), end-to-end latency by 80.8 - 94.2%, and energy consumption by 80.9 - 94.2%. In addition, we successfully deployed LifeLearner on two edge devices and a microcontroller unit, thereby enabling efficient CL on resource-constrained platforms where it would be impractical to run SOTA methods and the far-reaching deployment of adaptable CL in a ubiquitous manner. Code is available at https://github.com/theyoungkwon/LifeLearner.
Text
3625687.3625804
- Version of Record
More information
Published date: 12 November 2023
Venue - Dates:
21st ACM Conference on Embedded Networked Sensors Systems, SenSys 2023, , Istanbul, Turkey, 2023-11-13 - 2023-11-15
Keywords:
continual learning, edge computing, latent replay, meta learning, microcontrollers, on-device training, product quantization
Identifiers
Local EPrints ID: 491031
URI: http://eprints.soton.ac.uk/id/eprint/491031
PURE UUID: 3bbf0691-ff65-4065-9689-9cbdb760d31e
Catalogue record
Date deposited: 11 Jun 2024 16:41
Last modified: 14 Jun 2024 17:23
Export record
Altmetrics
Contributors
Author:
Young D. Kwon
Author:
Jagmohan Chauhan
Author:
Hong Jia
Author:
Stylianos I. Venieris
Author:
Cecilia Mascolo
Download statistics
Downloads from ePrints over the past year. Other digital versions may also be available to download e.g. from the publisher's website.
View more statistics