DeepWear: Adaptive Local Offloading for On-Wearable Deep Learning
Open accessThe DOI matches its Crossref record; 50 of the 51 references it cites were found in a registry or an index.
Abstract
Due to their on-body and ubiquitous nature, wearables can generate a wide range of unique sensor data creating countless opportunities for deep learning tasks. We propose DeepWear, a deep learning (DL) framework for wearable devices to improve the performance and reduce the energy footprint. DeepWear strategically offloads DL tasks from a wearable device to its paired handheld device through local network connectivity such as Bluetooth. Compared to the remote-cloud-based offloading, DeepWear requires no Internet connectivity, consumes less energy, and is robust to privacy breach. DeepWear provides various novel techniques such as context-aware offloading, strategic model partition, and pipelining support to efficiently utilize the processing capacity from nearby paired handhelds. Deployed as a user-space library, DeepWear offers developer-friendly APIs that are as simple as those in traditional DL libraries such as TensorFlow. We have implemented DeepWear on the Android OS and evaluated it on COTS smartphones and smartwatches with real DL models. DeepWear brings up to 5.08× and 23.0× execution speedup, as well as 53.5 and 85.5 percent energy saving compared to wearable-only and handheld-only strategies, respectively.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“In this manner, the complex DAG can be transformed into a linear and much simpler one, thus enabling a linear complexity solution for selecting the optimal partition to offload.”
From Convergence of Edge Computing and Deep Learning: A Comprehensive Survey (2020) · cites this paper as [174]
Citations
Cites
Going deeper with convolutions
Szegedy, C., Liu, W., Jia, Y., Sermanet, P., Reed, S., Anguelov, D., Erhan, D., Vanhoucke, V., & Rabinovich, A. (2015) · 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2015.7298594 · Cited by 48,562
MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications
Howard, A. G., Zhu, M., Chen, B., Kalenichenko, D., Wang, W., Weyand, T., Andreetto, M., & Adam, H. (2017) · arXiv · doi:10.48550/arxiv.1704.04861 · Cited by 19,015 · cited 7×
“MNIST [41] digit recognition grayscale image 15M MobileNet [30] image classification rgb image 580M GoogLeNet [42] image classification rgb image 2G LSTM-HAR [43] activity recognition mobile sensor 180M DeepSense [44] activity recognition mobile sensor 550M TextRNN [45] document classification word vectors 11M DeepEar…”
BackgroundInfluentialSelf-citationSequence to Sequence Learning with Neural Networks
Sutskever, I., Vinyals, O., & Le, Q. V. (2014) · arXiv · doi:10.48550/arxiv.1409.3215 · Cited by 17,766 · cited 1×
“For example, the LSTM model used for language modeling can also be applied to problems such as machine translation [47], question answering [48], and handwriting generation [49].”
MethodWaveNet: a generative model for raw audio
Oord, A. V. D., Dieleman, S., Zen, H., Simonyan, K., Vinyals, O., Graves, A., Kalchbrenner, N., Senior, A., & Kavukcuoglu, K. (2016) · arXiv · doi:10.48550/arxiv.1609.03499 · Cited by 4,911 · cited 1×
“…[42] image classification rgb image 2G LSTM-HAR [43] activity recognition mobile sensor 180M DeepSense [44] activity recognition mobile sensor 550M TextRNN [45] document classification word vectors 11M DeepEar [6] emotion recognition raw sound 9M WaveNet [46] speech recognition mfcc features 3.8G 1.”
BackgroundGenerating Sequences With Recurrent Neural Networks
Graves, A. (2013) · arXiv · doi:10.48550/arxiv.1308.0850 · Cited by 2,894 · cited 1×
“For example, the LSTM model used for language modeling can also be applied to problems such as machine translation [47], question answering [48], and handwriting generation [49].”
MethodMAUI: making smartphones last longer with code offload
Cuervo, E., Balasubramanian, A., Cho, D.-K., Wolman, A., Saroiu, S., Chandra, R., & Bahl, P. (2010) · Proceedings of the 8th international conference on Mobile systems, applications, and services · doi:10.1145/1814433.1814441 · Cited by 2,281 · cited 4×
“MAUI [13] designs an enhanced offloading decision mechanism that makes predictions for every single function invocation separately and considers the entire application when choosing which function to offload.”
BackgroundInfluentialOptimizing FPGA-based Accelerator Design for Deep Convolutional Neural Networks
Zhang, C., Li, P., Sun, G., Guan, Y., Xiao, B., & Cong, J. (2015) · Proceedings of the 2015 ACM/SIGDA International Symposium on Field-Programmable Gate Arrays · doi:10.1145/2684746.2689060 · Cited by 2,158 · cited 1×
“Some other efforts such as [31], [32], [33], [34] aimed at building customized hardware accelerators for DL or other machine learning tasks.”
BackgroundEIE: efficient inference engine on compressed deep neural network
Han, S., Liu, X., Mao, H., Pu, J., Pedram, A., Horowitz, M. A., & Dally, W. J. (2016) · ACM SIGARCH Computer Architecture News · doi:10.1145/3007787.3001163 · Cited by 1,937 · cited 1×
“Some other efforts such as [31], [32], [33], [34] aimed at building customized hardware accelerators for DL or other machine learning tasks.”
BackgroundSelf-citation(2011) · doi:10.1145/1966445.1966473 · Cited by 1,870
Exploiting Linear Structure Within Convolutional Networks for Efficient Evaluation
Denton, R., Zaremba, W., Bruna, J., LeCun, Y., & Fergus, R. (2014) · arXiv · doi:10.48550/arxiv.1404.0736 · Cited by 1,414 · cited 2×
“Besides, various model compression techniques [35], [3], [17], [18], [19], [36] have been proposed for accelerating the DL task and reducing its energy consumption.”
BackgroundMethod
Cited by
Convergence of Edge Computing and Deep Learning: A Comprehensive Survey
Wang, X., Han, Y., Leung, V. C. M., Niyato, D., Yan, X., & Chen, X. (2020) · IEEE Communications Surveys & Tutorials · doi:10.1109/comst.2020.2970550 · Cited by 1,498
“To efficiently determine partial offloading decisions, DeepWear first prunes the DAG by keeping only the computation-intensive nodes, and then grouping the repeated sub-DAGs.”
MethodInfluentialEdge AI: On-Demand Accelerating Deep Neural Network Inference via Edge Computing
Li, E., Zeng, L., Zhou, Z., & Chen, X. (2020) · IEEE Transactions on Wireless Communications · doi:10.1109/twc.2019.2946140 · Cited by 929
A Survey on Wearable Technology: History, State-of-the-Art and Current Challenges
Ometov, A., Shubina, V., Klus, L., Skibińska, J., Saafi, S., Pascacio, P., Flueratoru, L., Gaibor, D. Q., Chukhno, N., Chukhno, O., et al. (2021) · Computer Networks · doi:10.1016/j.comnet.2021.108074 · Cited by 506
Edge Intelligence: Empowering Intelligence to the Edge of Network
Xu, D., Li, T., Li, Y., Su, X., Tarkoma, S., Jiang, T., Crowcroft, J., & Hui, P. (2021) · Proceedings of the IEEE · doi:10.1109/jproc.2021.3119950 · Cited by 272
Edge Intelligence: Architectures, Challenges, and Applications
Xu, D., Li, T., Li, Y., Su, X., Tarkoma, S., Jiang, T., Crowcroft, J., & Hui, P. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2003.12172 · Cited by 120
Deep Learning on Computational-Resource-Limited Platforms: A Survey
Chen, C., Zhang, P., Zhang, H., Dai, J., Yi, Y., Zhang, H., & Zhang, Y. (2020) · Mobile Information Systems · doi:10.1155/2020/8454327 · Cited by 227
“Figure 1 shows typical applications of computational-resource-limited DL in the smart sensing context, including self-driving [25, 26], artificial intelligence APPs of smartphones [27], health/homecare robots [28–31], and intelligent wearable devices [32].”
BackgroundEdgeMoE: Empowering Sparse Large Language Models on Mobile Devices
Yi, R., Guo, L., Wei, S., Zhou, A., Wang, S., & Xu, M. (2025) · IEEE Transactions on Mobile Computing · doi:10.1109/tmc.2025.3546466 · Cited by 226
“One is at system level, e.g., by exploiting heterogeneous processors [21], [40], [45], [53], cache [66], [90], generating high-performance GPUs kernels [59], or adaptive offloading [54], [89].”
Self-citationOn-Demand Deep Model Compression for Mobile Devices: A Usage-Driven Model Selection Framework
Liu, S., Lin, Y., Zhou, Z., Nan, K., Liu, H., & Du, J. (2018) · Proceedings of the 16th Annual International Conference on Mobile Systems, Applications, and Services · doi:10.1145/3210240.3210337 · Cited by 217
Liang, Y., Samtani, S., Guo, B., & Yu, Z. (2020) · IEEE Internet of Things Journal · doi:10.1109/jiot.2020.3004077 · Cited by 214
Characterizing Impacts of Heterogeneity in Federated Learning upon Large-Scale Smartphone Data
Yang, C., Wang, Q., Xu, M., Chen, Z., Bian, K., Liu, Y., & Liu, X. (2021) · Proceedings of the Web Conference 2021 · doi:10.1145/3442381.3449851 · Cited by 135
“We are aware of learning-based approaches [51] to obtain the ondevice ML performance, but our empirical efforts show that these approaches are not precise enough for on-device training tasks.”
MethodSelf-citationBaccour, E., Mhaisen, N., Abdellatif, A. A., Erbad, A., Mohamed, A., Hamdi, M., & Guizani, M. (2022) · IEEE Communications Surveys & Tutorials · doi:10.1109/comst.2022.3200740 · Cited by 186
“For example, the DeepWear [225] approach splits the DNN into two sub-models that are separately computed on a wearable and a mobile device.”
MethodA First Look at Deep Learning Apps on Smartphones
Xu, M., Liu, J., Liu, Y., Lin, F. X., Liu, Y., & Liu, X. (2019) · The World Wide Web Conference · doi:10.1145/3308558.3313591 · Cited by 182
“, model compression [50, 59, 67, 74, 97], hardware customizations [45, 47, 57, 103], and cloud offloading [61, 66, 99, 105].”
BackgroundSelf-citationDNNOff: Offloading DNN-Based Intelligent IoT Applications in Mobile Edge Computing
Chen, X., Li, M., Zhong, H., Ma, Y., & Hsu, C. (2021) · IEEE Transactions on Industrial Informatics · doi:10.1109/tii.2021.3075464 · Cited by 155
“Second, the DNN programDP (li, 2, 4) 285 runs on the Node A; Pipe(1) and l1 are skipped, as CurrentL 286 is 2; Pipe(2), l2, Pipe(3), and l3 are executed in sequence, as 287 config[2] and config[3] are both Node A; Pipe(3) is executed 288 and the remote service is invoked, as config[4] is Node B.”
BackgroundAsynchronous Mobile-Edge Computation Offloading: Energy-Efficient Resource Management
You, C., Zeng, Y., Zhang, R., & Huang, K. (2018) · IEEE Transactions on Wireless Communications · doi:10.1109/twc.2018.2868710 · Cited by 132
Towards Energy Efficiency in the Internet of Wearable Things: A Systematic Review
Qaim, W. B., Ometov, A., Molinaro, A., Lener, I., Campolo, C., Lohan, E. S., & Nurmi, J. (2020) · IEEE Access · doi:10.1109/access.2020.3025270 · Cited by 131
“[113] proposed a generic deep learning framework for wearables to achieve improvements in their performance and energy consumption.”
BackgroundMelon: breaking the memory wall for resource-efficient on-device machine learning
Wang, Q., Xu, M., Jin, C., Dong, X., Yuan, J., Jin, X., Huang, G., Liu, Y., & Liu, X. (2022) · International Conference on Mobile Systems, Applications, and Services · doi:10.1145/3498361.3538928 · Cited by 100
Self-citationA Survey of Deep Learning on CPUs: Opportunities and Co-Optimizations
Mittal, S., Rajput, P., & Subramoney, S. (2021) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2021.3071762 · Cited by 85
On-device Training: A First Overview on Existing Systems
Zhu, S., Voigt, T., Rahimian, F., & Ko, J. (2024) · ACM Transactions on Sensor Networks · doi:10.1145/3696003 · Cited by 83
“Researchershavepresentedextensiveworksusingsuchan‘inference-embedded’systemarchitecture[23,56,66,120,126– 129, 131, 133].”
Grzesik, P., & Mrozek, D. (2024) · Electronics · doi:10.3390/electronics13030640 · Cited by 75
“In order to enhance the performance of healthcare-related wearables, Xu et al. [81] proposed “DeepWear”, which is a deep learning framework dedicated to wearable devices that allows them to offload some of the computational tasks to edge devices such as smart-phones.”
BackgroundMandheling: mixed-precision on-device DNN training with DSP offloading
Xu, D., Xu, M., Wang, Q., Wang, S., Ma, Y., Huang, K. R., Huang, G., Jin, X., & Liu, X. (2022) · ACM Annual International Conference on Mobile Computing and Networking (MobiCom) · doi:10.1145/3495243.3560545 · Cited by 75
“example, the model could be pre-trained on the cloud with public data and fine-tuned or personalized on each device [76, 79, 80].”
Self-citationAcheampong, A., Zhang, Y., Xu, X., & Kumah, D. A. (2022) · Computer Modeling in Engineering & Sciences · doi:10.32604/cmes.2022.021394 · Cited by 71
Joint Optimization With DNN Partitioning and Resource Allocation in Mobile Edge Computing
Dong, C., Hu, S., Chen, X., & Wen, W. (2021) · IEEE Transactions on Network and Service Management · doi:10.1109/tnsm.2021.3116665 · Cited by 70
A Comprehensive Benchmark of Deep Learning Libraries on Mobile Devices
Zhang, Q., Li, X. Y., Che, X., Ma, X., Zhou, A., Xu, M., Wang, S., Ma, Y., & Liu, X. (2022) · International World Wide Web Conference (WWW) · doi:10.1145/3485447.3512148 · Cited by 71
Self-citationCloud Versus Edge Deployment Strategies of Real-Time Face Recognition Inference
Koubaa, A., Ammar, A., Kanhouch, A., & AlHabashi, Y. (2022) · IEEE Transactions on Network Science and Engineering · doi:10.1109/tnse.2021.3055835 · Cited by 69
“, 2019 [21] Digit recognition Image classification Activity recognition Document classification Emotion recognition Speech recognition MobileNet [30] GoogLeNet [38] LSTM-HAR [39] DeepSense [40] TextRNN [41] DeepEar [42] WaveNet [43] COTS smartphones and smartwatches Context-aware offloading of DL tasks from wearable t…”
BackgroundHuang, H., Ye, Q., & Zhou, Y.-T. (2021) · IEEE Transactions on Network Science and Engineering · doi:10.1109/tnse.2021.3115054 · Cited by 65
“designed a deep learning based offloading scheme for wearable devices [28].”
BackgroundZhou, P., Xu, H., Lee, L., Fang, P., & Hui, P. (2022) · Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies · doi:10.1145/3534585 · Cited by 63
EPASS360: QoE-Aware 360-Degree Video Streaming Over Mobile Devices
Zhang, Y., Guan, Y., Bian, K., Liu, Y., Tuo, H., Song, L., & Li, X. (2020) · IEEE Transactions on Mobile Computing · doi:10.1109/tmc.2020.2978187 · Cited by 55
“if not inherit unaffordable complexity, can be employed to tackle the on-device issues [18], [19].”
BackgroundLattanzi, E., Donati, M., & Freschi, V. (2022) · Sensors · doi:10.3390/s22072637 · Cited by 54
Nho, Y., Ryu, S., & Kwon, D. (2021) · IEEE Sensors Journal · doi:10.1109/jsen.2021.3054394 · Cited by 49
An efficient DNN splitting scheme for edge-AI enabled smart manufacturing
Gauttam, H., Pattanaik, K. K., Bhadauria, S., Nain, G., & Prakash, P. B. (2023) · Journal of Industrial Information Integration · doi:10.1016/j.jii.2023.100481 · Cited by 43
A Decade of Video Analytics at Edge: Training, Deployment, Orchestration, and Platforms
Bai, T., Zhao, H., Huang, L., Wang, Z., Kim, D. I., & Nallanathan, A. (2025) · IEEE Communications Surveys & Tutorials · doi:10.1109/comst.2025.3563377 · Cited by 43
“Hence, GRAMI was utilized for structure linearization, to enable the existing partitioning method [171].”
Scission: Performance-driven and Context-aware Cloud-Edge Distribution of Deep Neural Networks
Lockhart, L., Harvey, P., Imai, P., Willis, P., & Varghese, B. (2020) · IEEE/ACM International Conference on Utility and Cloud Computing (UCC) · doi:10.1109/ucc48980.2020.00044 · Cited by 43
“DeepWear similarly uses a similar approach to train prediction models to estimate latency and energy consumption of four popular layer types and their parameter combinations across a wearable and its paired device [30].”
MethodEdge-AI: A systematic review on architectures, applications, and challenges
Gauttam, H., Nain, G., Pattanaik, K. K., & Mendes, P. (2025) · Journal of Network and Computer Applications · doi:10.1016/j.jnca.2025.104375 · Cited by 42
Baccarelli, E., Scarpiniti, M., Momenzadeh, A., & Ahrabi, S. S. (2021) · IEEE Access · doi:10.1109/access.2021.3058021 · Cited by 43
“Overall, like our proposal, all the contributions in [17], [18] and [19] consider the problem of the optimized execution of DNNs over distributed execution platforms under various system constraints, but, unlike our paper, they do not afford the problem of the energy-saving dynamic joint scaling of…”
BackgroundMethodAn Emerging Decentralized Services Computing Paradigm for Big Data Governance: A Position Paper
Liu, X., Sun, X., & Huang, G. (2019) · IEEE Transactions on Services Computing · doi:10.1109/tsc.2019.2951558 · Cited by 41
“Additionally, some new techniques, such as local offloading [39], serverless analytics [40], and modeless inference [41] are worth exploring.”
BackgroundSelf-citationLiu, S., Du, J., Nan, K., Zhou, Z., Liu, H., Wang, Z., & Lin, Y. (2020) · IEEE Transactions on Mobile Computing · doi:10.1109/tmc.2020.2999956 · Cited by 41
Deep Learning for Clinical Decision Support Systems: A Review from the Panorama of Smart Healthcare
Kumar, E. S., & Jayadev, P. S. (2019) · Studies in big data · doi:10.1007/978-3-030-33966-1_5 · Cited by 40
6G-Empowered Offloading for Realtime Applications in Multi-Access Edge Computing
Huang, H., Ye, Q., & Zhou, Y.-T. (2022) · IEEE Transactions on Network Science and Engineering · doi:10.1109/tnse.2022.3188921 · Cited by 39
“In work [23], the authors proposed a holistic deep learning-based offloading model in wearable devices to offer local offloading services for wearable deep learning applications.”
BackgroundCrowdHMT: Crowd Intelligence With the Deep Fusion of Human, Machine, and IoT
Guo, B., Liu, Y., Liu, S., Yu, Z., & Zhou, X. (2022) · IEEE Internet of Things Journal · doi:10.1109/jiot.2022.3194726 · Cited by 38
“Deepware [64] uses Monsoon Power Meter to quantify the energy consumption of deep models in smartphones, and investigates the energy consumption model”
MethodStochastic Cumulative DNN Inference With RL-Aided Adaptive IoT Device-Edge Collaboration
Qu, K., Zhuang, W., Wu, W. R., Li, M., Shen, X., Li, X., & Shi, W. (2023) · IEEE Internet of Things Journal · doi:10.1109/jiot.2023.3280746 · Cited by 38
Nakhkash, M., Gia, T. N., Azimi, I., Anzanpour, A., Rahmani, A. M., & Liljeberg, P. (2019) · IEEE International Conference on Omni-layer Intelligent Systems (COINS) · doi:10.1145/3312614.3312632 · Cited by 37
Samikwa, E., Maio, A. D., & Braun, T. I. (2022) · IEEE Consumer Communications and Networking Conference · doi:10.1109/ccnc49033.2022.9700550 · Cited by 36
Calibration-Aided Edge Inference Offloading via Adaptive Model Partitioning of Deep Neural Networks
Pacheco, R. G., Couto, R. S., & Simeone, O. (2021) · International Conference on Communications · doi:10.1109/icc42927.2021.9500760 · Cited by 35
“wearables to execute DNN inference locally [4].”
Deep reinforcement learning in edge networks: Challenges and future directions
Hazra, A., Tummala, V. M. R., Mazumdar, N., Sah, D. K., & Adhikari, M. (2024) · Physical Communication · doi:10.1016/j.phycom.2024.102460 · Cited by 34
SEE: Scheduling Early Exit for Mobile DNN Inference during Service Outage
Wang, Z., Bao, W., Yuan, D., Ge, L., Tran, N. H. K., & Zomaya, A. Y. (2019) · ACM International Conference on Modeling, Analysis and Simulation of Wireless and Mobile Systems (MSWiM) · doi:10.1145/3345768.3355917 · Cited by 34
Arif, M., & Rashid, M. (2025) · Computers, materials & continua/Computers, materials & continua (Print) · doi:10.32604/cmc.2025.062819 · Cited by 33
Context-Aware Edge-Based AI Models for Wireless Sensor Networks—An Overview
Al-Saedi, A. A., Boeva, V., Casalicchio, E., & Exner, P. (2022) · Sensors · doi:10.3390/s22155544 · Cited by 33
A Comprehensive Deep Learning Library Benchmark and Optimal Library Selection
Zhang, Q., Che, X., Chen, Y., Ma, X., Xu, M., Dustdar, S., Liu, X., & Wang, S. (2023) · IEEE Transactions on Mobile Computing · doi:10.1109/tmc.2023.3301973 · Cited by 32
Self-citationPartitioning of CNN Models for Execution on Fog Devices
Dey, S., Mukherjee, A., Pal, A., & Balamuralidhar, P. (2018) · doi:10.1145/3277893.3277899 · Cited by 32
Qiu, J., Yan, X., Wang, W. W., Wei, W., & Fang, K. (2021) · IEEE Journal of Biomedical and Health Informatics · doi:10.1109/jbhi.2021.3137334 · Cited by 31
“Xu et al. [29] optimized Neurosurgeon and applied it in embedded wearable devices to improve the performance and reduce the energy footprint.”
BackgroundHarnessing the Computing Continuum for Urgent Science
Balouek‐Thomert, D., Rodero, I., & Parashar, M. (2020) · ACM SIGMETRICS Performance Evaluation Review · doi:10.1145/3439602.3439618 · Cited by 29
“to powerful multi-node cloud-based engines [26, 27].”
A Co-Scheduling Framework for DNN Models on Mobile and Edge Devices with Heterogeneous Hardware
Xu, Z., Yang, D., Yin, C., Tang, J., Wang, Y., & Xue, G. (2021) · IEEE Transactions on Mobile Computing · doi:10.1109/tmc.2021.3107424 · Cited by 29
“Unlike [13], [25], [62], [66], [68], which addressed the problem of offloading DNN models (or part of models) to the cloud, we, however, focus on the on-device inference without the help of cloud servers, and consider a mathematically different problem of scheduling tasks on heterogeneous hardware.”
BackgroundOnline Learning for Orchestration of Inference in Multi-user End-edge-cloud Networks
Shahhosseini, S., Seo, D., Kanduri, A., Hu, T., Lim, S.-S., Donyanavard, B., Rahmani, A. M., & Dutt, N. D. (2022) · ACM Transactions on Embedded Computing Systems · doi:10.1145/3520129 · Cited by 28
“Generally, DL models are partitioned according to the compute cost of model layers and required bandwidth for each layer to be distributed among the end-edge-cloud [15, 16, 34, 46].”
MethodManzoor, S., Kim, E., Joo, S.-H., Bae, S.-H., In, G.-G., Joo, K.-J., Choi, J.-H., & Kuc, T. (2022) · IEEE Access · doi:10.1109/access.2022.3190538 · Cited by 28
Horizontally Distributed Inference of Deep Neural Networks for AI-Enabled IoT
Rodríguez, I., Campos, C., & Fdez‐Riverola, F. (2023) · Sensors · doi:10.3390/s23041911 · Cited by 28
Shi, L., Xu, Z., Sun, Y., Shi, Y., Fan, Y., & Ding, X. (2021) · Peer-to-Peer Networking and Applications · doi:10.1007/s12083-021-01223-1 · Cited by 28
Boosting DNN Cold Inference on Edge Devices
Yi, R., Cao, T., Zhou, A., Ma, X., Wang, S., & Xu, M. (2023) · International Conference on Mobile Systems, Applications, and Services · doi:10.1145/3581791.3596842 · Cited by 27
“, by exploiting heterogeneous processors [20, 25, 28, 34], cache [45, 61, 67], generating high-performance GPUs kernels [40], or adaptive offloading [35, 65].”
BackgroundSelf-citationBaccour, E., Erbad, A., Mohamed, A. M. S., Hamdi, M., & Guizani, M. (2024) · Future Generation Computer Systems · doi:10.1016/j.future.2024.01.021 · Cited by 27
Joint Batching and Scheduling for High-Throughput Multiuser Edge AI With Asynchronous Task Arrivals
Cang, Y., Chen, M., & Huang, K. (2024) · IEEE Transactions on Wireless Communications · doi:10.1109/twc.2024.3404811 · Cited by 25
Park, J., Lim, W.-S., Kim, D., & Lee, J. (2023) · Engineering Applications of Artificial Intelligence · doi:10.1016/j.engappai.2023.106543 · Cited by 25
Virtual Connection: Selective Connection System for Energy-Efficient Wearable Consumer Electronics
Hong, G., & Shin, D. (2020) · IEEE Transactions on Consumer Electronics · doi:10.1109/tce.2020.3035689 · Cited by 24
“puting resources [8]–[12] to its companion wearable devices.”
BackgroundLi, X., Qin, Y., Zhou, H., & Zhang, Z. (2021) · Transactions on Emerging Telecommunications Technologies · doi:10.1002/ett.4263 · Cited by 24
“Based on the wearable devices, Xu et al 21 proposed a developer-friendly framework named DeepWear to strategically offload DL-based tasks from a wearable device to its paired handheld device by local network communication technique.”
MethodPacheco, R. G., Couto, R. S., & Simeone, O. (2023) · Journal of Network and Computer Applications · doi:10.1016/j.jnca.2023.103679 · Cited by 24
“The partitioning layer is chosen by employing an optimization problem for a given objective, such as reducing the inference time [9] or saving energy on the edge device [10].”
MethodAn Integrated Cloud-Edge-Device Adaptive Deep Learning Service for Cross-Platform Web
Huang, Y., Qiao, X., Tang, J., Ren, P., Liu, L., Pu, C., & Chen, J. (2021) · IEEE Transactions on Mobile Computing · doi:10.1109/tmc.2021.3122279 · Cited by 21
“computation at a modest loss in accuracy as a trade-off [15], [16], [17].”
BackgroundTowards Ubiquitous Learning: A First Measurement of On-Device Training Performance
Cai, D., Wang, Q., Liu, Y., Liu, Y., Wang, S., & Xu, M. (2021) · International Workshop on Embedded and Mobile Deep Learning (EMDL) · doi:10.1145/3469116.3470009 · Cited by 21
Self-citationGenerative AI-Aided Multimodal Parallel Offloading for AIGC Metaverse Service in IoT Networks
Zeng, W., Zheng, J., Gao, L. C., Niu, J., Ren, J., Wang, H., Cao, R., & Ji, S. (2025) · IEEE Internet of Things Journal · doi:10.1109/jiot.2025.3535623 · Cited by 20
“A deep learning (DL) framework for wearable devices was proposed to enhance performance and minimize energy consumption by offload-ing tasks to paired handhelds via local connectivity [20].”
Towards scalable and efficient Deep-RL in edge computing: A game-based partition approach
Dai, H., Wu, J., Wang, Y., & Xu, C. (2022) · Journal of Parallel and Distributed Computing · doi:10.1016/j.jpdc.2022.06.006 · Cited by 19
“Deep reinforcement learning (Deep-RL) [39,42], which integrates deep neural network (DNN) into a framework of reinforcement learning (RL) to help software agents learn how to reach their goals, has been widely deployed on various mobile or stationary devices, such as smartphones, wearables, smart cameras, and other in…”
BackgroundEnergy-Aware Deep Learning on Resource-Constrained HardwarePreprint
Millar, J., Haddadi, H., & Madhavapeddy, A. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2505.12523 · Cited by 19
Integrated Communication and Computing Maritime Networks Design for Green Metaverse
Yang, T., Cui, Z., Peng, C., Wu, J., Liu, F., & Yang, Y. (2023) · IEEE Wireless Communications · doi:10.1109/mwc.017.2300119 · Cited by 17
“If a work device does not have the necessary information for the calculation, it can request work devices [10].”
BackgroundEnergy Efficient and Real-Time Remote Sensing in AI-Powered Drone
Kim, B., Jung, J., Min, H., & Heo, J. (2021) · Mobile Information Systems · doi:10.1155/2021/6650053 · Cited by 17
“DeepWear was proposed to decide on deep learning tasks for wearable devices [19].”
BackgroundFUNOff: Offloading Applications At Function Granularity for Mobile Edge Computing
Chen, X., Li, M., Zhong, H., Chen, X., Ma, Y., & Hsu, C. (2023) · IEEE Transactions on Mobile Computing · doi:10.1109/tmc.2023.3240741 · Cited by 16
“Although the prior studies [20], [21], [22] can be extended to the scenario of MEC, they lack enough effectiveness, since they only divide an application into two parts and deploy them on a mobile device and a remote server.”
BackgroundInfluentialLattanzi, E., Calisti, L., & Capellacci, P. (2023) · Pervasive and Mobile Computing · doi:10.1016/j.pmcj.2023.101848 · Cited by 16
Khlaisang, J., & Koraneekij, P. (2024) · The Electronic Journal of e-Learning · doi:10.34190/ejel.22.8.3505 · Cited by 16
“…technology media and learning challenges (Shahmohammadi et al., 2017; Babic, Gaspar, and Satala, 2018), as well as flexible learning and the use of a learning system that can be accessed anywhere and anytime (Al-Soh and Zualkernan, 2017; Royal Thai Government Gazette, 2018; Xu et al., 2020).”
BackgroundThe Case for Adaptive Deep Neural Networks in Edge Computing
McNamee, F., Dustdar, S., Kilpatrick, P., Shi, W., Spence, I. T. A., & Varghese, B. (2021) · IEEE ... International Conference on Cloud Computing · doi:10.1109/cloud53861.2021.00017 · Cited by 16
“These methods include, estimation-based techniques [6], [13], [14], structural modification-based techniques [15], [16], and measurement-based techniques [17].”
MethodDNNPipe: Dynamic programming-based optimal DNN partitioning for pipelined inference on IoT networks
Seo, W., Kim, S., & Hong, S. (2025) · Journal of Systems Architecture · doi:10.1016/j.sysarc.2025.103462 · Cited by 14
AoI-Aware Incentive Mechanism for UAV-Assisted Mobile Crowdsensing: A Contract-Theoretic Approach
Guo, Y., Chen, Y., Li, H., Wu, Y., & Huang, J. (2025) · IEEE Transactions on Mobile Computing · doi:10.1109/tmc.2025.3604073 · Cited by 14
“Mobile crowdsensing (MCS), which utilizes mobile devices equipped with sensors to collect data, has emerged as a transformative approach [1], [2], [3], [4].”
Computation offloading technique for energy efficiency of smart devices
Ko, J., Choi, Y., & Paul, R. (2021) · Journal of Cloud Computing Advances Systems and Applications · doi:10.1186/s13677-021-00260-8 · Cited by 14
Collaborative Intelligent Cross-Camera Video Analytics at Edge: Opportunities and Challenges
Pasandi, H. B., & Nadeem, T. (2019) · International Workshop on Challenges in Artificial Intelligence and Machine Learning for Internet of Things (AIChallengeIoT) · doi:10.1145/3363347.3363360 · Cited by 13
Understanding the Performance of Task Offloading for Wearables in a Two-Tier Edge Architecture
Qaim, W. B., Ometov, A., Campolo, C., Molinaro, A., Lohan, E. S., & Nurmi, J. (2021) · International Congress on Ultra Modern Telecommunications and Control Systems and Workshops (ICUMT) · doi:10.1109/icumt54235.2021.9631613 · Cited by 13
Mercati, P., & Bhat, G. (2024) · IEEE Design and Test · doi:10.1109/mdat.2024.3432862 · Cited by 11
“Another class of edge-cloud collaboration approaches performs initial processing on the device to minimize data transmission [176], [177], [178], [179], [180].”
Artificial intelligence-enabled wearable microgrids for self-sustained energy management
Ding, S., Bian, Y., Saha, T., Khan, M. I., Chang, A., Yin, L., Xu, S. H., & Wang, J. (2025) · Nature Reviews Electrical Engineering · doi:10.1038/s44287-025-00206-1 · Cited by 11
Evaluating policy-driven adaptation on the Edge-to-Cloud Continuum
Balouek‐Thomert, D., Rodero, I., & Parashar, M. (2021) · IEEE/ACM HPC for Urgent Decision Making (UrgentHPC) · doi:10.1109/urgenthpc54802.2021.00007 · Cited by 11
Velasco-Montero, D., Goossens, B., Fernández‐Berni, J., Rodriguez-Vazquez, A., & Philips, W. R. (2023) · IEEE Access · doi:10.1109/access.2023.3264828 · Cited by 11
“Multi-node edge-edge solutions can also optimize network execution by applying layer fusion, data parallelism, or network partition over clusters of edge devices [18], [19], [30], [31], [32], [33].”
BackgroundSCL: A sustainable deep learning solution for edge computing ecosystem in smart manufacturing
Gauttam, H., Pattanaik, K. K., Bhadauria, S., & Nain, G. (2024) · Journal of Industrial Information Integration · doi:10.1016/j.jii.2024.100703 · Cited by 11
Energy-efficient Collaborative Sensing: Learning the Latent Correlations of Heterogeneous Sensors
Liang, Y., Wang, X., Yu, Z., Guo, B., Zheng, X., & Samtani, S. (2021) · ACM Transactions on Sensor Networks · doi:10.1145/3448416 · Cited by 11
“Via edge intelligence, the complex and energy-intensive deep neural networks can be partitioned into tiny subtasks, and be distributively performed on neighboring devices or edges [20, 66].”
BackgroundEdge resource slicing approaches for latency optimization in AI-edge orchestration
Chandrika, P. K., Mekala, M. S., & Srivastava, G. (2022) · Cluster Computing · doi:10.1007/s10586-022-03817-7 · Cited by 11
“Furthermore, workload computation can process and compute as per the Dynamic Acyclic Graph (DAG) mechanism [44], where every ED depicts a layer, and the edge represents the flow of data between such layers.”
MethodLoADPart: Load-Aware Dynamic Partition of Deep Neural Networks for Edge Offloading
Liu, H., Zheng, W., Li, L., & Guo, M. (2022) · IEEE International Conference on Distributed Computing Systems (ICDCS) · doi:10.1109/icdcs54860.2022.00053 · Cited by 11
“Therefore, partial offloading [4], [5] is presented with each DNN partitioned and distributed across the edge server and user-end device to enable collaborative inference.”
BackgroundHybrid mobile vision for emerging applications
Wu, N., Lin, F. X., Qian, F., & Han, B. (2022) · IEEE Workshop on Mobile Computing Systems and Applications (WMCSA) · doi:10.1145/3508396.3512876 · Cited by 9
“While selective offloading [17, 43] is not a completely new idea, a challenge here is to strategically partition the workload at the appropriate granularity (e.”
BackgroundSelf-citationNestQuant: Post-Training Integer-Nesting Quantization for On-Device DNN
Xie, J., Ding, C., Li, X., Ren, S., Li, Y., & Lu, Z. (2025) · IEEE Transactions on Mobile Computing · doi:10.1109/tmc.2025.3582583 · Cited by 8
“O N-DEVICE deploying deep neural network (DNN) models has become mainstream providing AI-powered deep learning (DL) services [1], [2], [3], [4], [5], [6], [7] on ubiquitous Internet of Things (IoT) devices, such as smartphones, wear-ables, and advanced driving systems, etc.”
Wang, Y., Yu, Z., Liu, S., Zhou, Z., & Guo, B. (2023) · Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies · doi:10.1145/3580815 · Cited by 8
Localized Knowledge Distillation Helps IoT Devices Provide High-performance Visual Services
Yu, J., Ding, C., Li, H., Jin, Y., & Li, Y. (2023) · IEEE International Conference on Web Services (ICWS) · doi:10.1109/icws60048.2023.00036 · Cited by 7
Embedded Distributed Inference of Deep Neural Networks: A Systematic Review
Peccia, F. N., & Bringmann, O. (2024) · arXiv (Cornell University) · doi:10.48550/arxiv.2405.03360 · Cited by 7
“Others [148] say they are interested in publishing their code, but the authors of this paper were not able to find any open implementation for them.”
BackgroundInfluentialSlimmable Encoders for Flexible Split DNNs in Bandwidth and Resource Constrained IoT Systems
Assine, J. S., Filho, J. C. S. S., Junior, E. A. D. V., & Levorato, M. (2023) · IEEE International Symposium on a World of Wireless, Mobile and Multimedia Networks (WoWMoM) · doi:10.1109/wowmom57956.2023.00014 · Cited by 7
ScissionLite: Accelerating Distributed Deep Neural Networks Using Transfer LayerPreprint
Ahn, H., Lee, M., Hong, C.-H., & Varghese, B. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2105.02019 · Cited by 7
Tian, J., Li, X., & Qin, X. (2024) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn60899.2024.10651115 · Cited by 6
“Each end device generates λ i ( t ) tasks in time slot t in the range [1 , 10] .”
BackgroundMethodSecoInfer: Secure DNN End-Edge Collaborative Inference Framework Optimizing Privacy and Latency
Yao, Y., Hou, J., Wu, G., Cheng, Y., Yuan, M., Luo, P., Wang, Z., & Li, X. (2024) · ACM Transactions on Sensor Networks · doi:10.1145/3694972 · Cited by 6
“DeepWear [44] can be considered a variation of DADS.”
Accelerating on-device DNN inference during service outage through scheduling early exit
Wang, Z., Bao, W., Yuan, D., Ge, L., Tran, N. H. K., & Zomaya, A. Y. (2020) · Computer Communications · doi:10.1016/j.comcom.2020.08.005 · Cited by 6
ScissionLite: Accelerating Distributed Deep Learning With Lightweight Data Compression for IIoT
Ahn, H., Lee, M., Seong, S., Na, G.-J., Chun, I., Varghese, B., & Hong, C.-H. (2024) · IEEE Transactions on Industrial Informatics · doi:10.1109/tii.2024.3413340 · Cited by 5
“The first is an estimation-based approach, in which the slicing point is determined based on an estimation of the performance of the layers on a target hardware platform [9], [14], [15], [16].”
MethodYang, L., Wang, S., Zhang, W., Jing, B., Yu, X., Tang, Z., & Wang, W. (2025) · Applied Sciences · doi:10.3390/app15179784 · Cited by 5
Partitioning multi-layer edge network for neural network collaborative computing
Li, Q., Zhou, M., Ren, T.-F., Jiang, C.-B., & Chen, Y. (2023) · EURASIP Journal on Wireless Communications and Networking · doi:10.1186/s13638-023-02284-x · Cited by 5
“The studies of [22–27] all discuss the two-portion neural network and collaborative computing scenarios, which are simpler than multipoint partitioning and require further design of decision-making for task partitioning in face of more devices collaborative computing.”
Background
Metadata
- Authors
- Xu, M., Qian, F., Zhu, M., Huang, F., Pushp, S., & Liu, X.
- Affiliations
- Peking University (CN)
- University of Minnesota (US)
- Korea Advanced Institute of Science and Technology (KR)
- Venue
- IEEE Transactions on Mobile Computing
- Published
- 2020-02-01
- Volume
- 19
- Issue
- 2
- Pages
- 314-330
- ISSN
- 1536-1233 (ISSN-L)
- 1558-0660
- 2161-9875
- OpenAlex
- W2909099954
- Cited by
- 138
- References
- 51
- Citations by year
- 2026: 10 · 2025: 22 · 2024: 13 · 2023: 17 · 2022: 24 · 2021: 33 · 2020: 14
- FWCI
- 10.40
- Type
- journal article
- Language
- en
- Open access
- green
- arxiv.org
- Licence
- Locations
- IEEE Transactions on Mobile Computing (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- arXiv (Cornell University)
- Funders
- National Key R&D Program (2018YFB1004800)
- National Natural Science Foundation of China (61725201)
- NSF (CCF-1629347)
- Basic Science Research Program (2017R1A2B3010504)
- Next-Generation Information Computing Development Program (2017M3C4A7065963)
- National Research Foundation of Korea
- Ministry of Science, ICT and Future Planning
- Topics
- Physical Sciences › Computer Science › Computer Networks and Communications › IoT and Edge/Fog Computing
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Context-Aware Activity Recognition Systems
- Physical Sciences › Engineering › Electrical and Electronic Engineering › Green IT and Sustainability
- Keywords
- wearable computing, model partitioning, pipelining, Bluetooth, energy efficiency, privacy preservation, smartwatches, smartphones, Android
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.