Deep Learning on Mobile Devices With Neural Processing Units
- 3 August 2023
- journal article
- research article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in Computer
- Vol. 56 (8) , 48-57
Abstract
Although deep neural networks (DNNs) can require less processing time when running on neural processing units (NPUs) instead of CPUs, their accuracy can degrade. We point out three research directions to improve DNN performance on mobile devices with NPUs.Keywords
Funding Information
- National Science Foundation (CCF-2125208, CCF-2215043)
This publication has 12 references indexed in Scilit:
- Edge Assisted Real-time Object Detection for Mobile Augmented Reality2019
- Dynamic Adaptive DNN Surgery for Inference Acceleration on the Edge2019
- MeDNN: A distributed mobile system with enhanced partition and deployment for large-scale DNNs2017
- Distributed Deep Neural Networks Over the Cloud, the Edge and End Devices2017
- NeurosurgeonACM SIGARCH Computer Architecture News, 2017
- Analyzing Classifiers: Fisher Vectors and Deep Neural Networks2016
- DeepX: A Software Accelerator for Low-Power Deep Learning Inference on Mobile Devices2016
- Early Implementation Experience with Wearable Cognitive Assistance Applications2015
- The Pascal Visual Object Classes Challenge: A RetrospectiveInternational Journal of Computer Vision, 2014
- MAUI2010