Dynamic Adaptive DNN Surgery for Inference Acceleration on the Edge
The DOI matches its Crossref record; 22 of the 23 references it cites were found in a registry or an index.
Abstract
Recent advances in deep neural networks (DNNs) have substantially improved the accuracy and speed of a variety of intelligent applications. Nevertheless, one obstacle is that DNN inference imposes heavy computation burden to end devices, but offloading inference tasks to the cloud causes transmission of a large volume of data. Motivated by the fact that the data size of some intermediate DNN layers is significantly smaller than that of raw input data, we design the DNN surgery, which allows partitioned DNN processed at both the edge and cloud while limiting the data transmission. The challenge is twofold: (1) Network dynamics substantially influence the performance of DNN partition, and (2) State-of-the-art DNNs are characterized by a directed acyclic graph (DAG) rather than a chain so that partition is greatly complicated. In order to solve the issues, we design a Dynamic Adaptive DNN Surgery (DADS) scheme, which optimally partitions the DNN under different network condition. Under the lightly loaded condition, DNN Surgery Light (DSL) is developed, which minimizes the overall delay to process one frame. The minimization problem is equivalent to a min-cut problem so that a globally optimal solution is derived. In the heavily loaded condition, DNN Surgery Heavy (DSH) is developed, with the objective to maximize throughput. However, the problem is NP-hard so that DSH resorts an approximation method to achieve an approximation ratio of 3. Real-world prototype based on self-driving car video dataset is implemented, showing that compared with executing entire the DNN on the edge and cloud, DADS can improve latency up to 6.45 and 8.08 times respectively, and improve throughput up to 8.31 and 14.01 times respectively.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Citations
Cites
Deep Residual Learning for Image Recognition
He, K., Zhang, X., Ren, S., & Sun, J. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.90 · Cited by 235,474
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
YOLO9000: Better, Faster, Stronger
Redmon, J., & Farhadi, A. (2017) · 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2017.690 · Cited by 19,892 · cited 4×
“For example, the input data size of tiny YOLOv2 [4] is 0 .”
BackgroundInfluentialInception-v4, Inception-ResNet and the Impact of Residual Connections on Learning
Szegedy, C., Ioffe, S., Vanhoucke, V., & Alemi, A. (2017) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v31i1.11231 · Cited by 4,636 · cited 1×
“6 shows the DAG of the pure inception v4 network [10] in Fig.”
BackgroundAn experimental comparison of min-cut/max- flow algorithms for energy minimization in vision
Boykov, Y., & Kolmogorov, V. (2004) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2004.60 · Cited by 4,577
MAUI: 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
Dynamic adaptive streaming over HTTP -: standards and design principles
Stockhammer, T. (2011) · Proceedings of the second annual ACM conference on Multimedia systems · doi:10.1145/1943552.1943572 · Cited by 1,221 · cited 1×
“This can be realized by a method similar to HTTP DASH [11].”
MethodDeep neural networks for small footprint text-dependent speaker verification
Variani, E., Lei, X., McDermott, E., Moreno, I. L., & Gonzalez-Dominguez, J. (2014) · 2014 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) · doi:10.1109/icassp.2014.6854363 · Cited by 1,035
Distributed Deep Neural Networks Over the Cloud, the Edge and End Devices
Teerapittayanon, S., McDanel, B., & Kung, H. T. (2017) · 2017 IEEE 37th International Conference on Distributed Computing Systems (ICDCS) · doi:10.1109/icdcs.2017.226 · Cited by 843
(2011) · Cited by 713
Cited by
Items 401–493 of 493
Recursive design for data-driven, self-adaptive IoT services
Frangoudis, P. A., Reisinger, M., & Dustdar, S. (2021) · IEEE International Symposium on Service-Oriented System Engineering (SOSE) · doi:10.1109/sose52839.2021.00009 · Cited by 1
“Distributing the computations of a neural network structure across nodes by model splitting and edge offloading is a topic recently receiving attention [27], [36], [35], [37], [38], [39], [40].”
BackgroundParticionamento de Redes Neurais Profundas com Saídas Antecipadas
Pacheco, R. G., & Couto, R. S. (2021) · Simpósio Brasileiro de Redes de Computadores e Sistemas Distribuídos (SBRC) · doi:10.5753/sbrc_estendido.2021.17168 · Cited by 1
“Neurosurgeon [Kang et al., 2017] e DADS [Hu et al., 2019] (Dynamic Adaptive DNN Surgery) propõem métodos de particionamento ótimo para reduzir o tempo de inferência.”
BackgroundReliable Edge Intelligence Using JPEG Progressive
Luo, H., Du, X., Dong, L., Su, G., & Chen, R. (2021) · Lecture notes in computer science · doi:10.1007/978-3-030-93176-6_30 · Cited by 1
An Efficient Scheme for Data Pattern Matching in IoT Networks
Ali, A. A., & Saraereh, O. A. (2022) · Computers, materials & continua/Computers, materials & continua (Print) · doi:10.32604/cmc.2022.025994 · Cited by 1
Overview of Edge Intelligence and Blockchain
Wang, X., Qiu, C., Ren, X., Xiong, Z., Leung, V. C. M., & Niyato, D. T. (2022) · Wireless networks · doi:10.1007/978-3-031-10186-1_2 · Cited by 1
DBM: Delay-sensitive Buffering Mechanism for DNN Offloading Services
Gao, G., Wu, L., Yang, Y., & Li, K. (2022) · Asia-Pacific Conference on Communications (APCC) · doi:10.1109/apcc55198.2022.9943625 · Cited by 1
“As a critical DNN task type, streaming tasks have gained increasing popularity [14], [23].”
BackgroundMethodStegEdge: Privacy protection of unknown sensitive attributes in edge intelligence via deception
Zhang, J., Zhang, W., & Xu, J. (2022) · Journal of Computer Security · doi:10.3233/jcs-220042 · Cited by 1
“We believe such selections are consistent with the practice in the relevant literature like [12,17].”
Zhang, Z., Yu, H., & Wang, F. (2022) · Journal of King Saud University - Computer and Information Sciences · doi:10.1016/j.jksuci.2022.12.013 · Cited by 1
“…exponentially and the evaluation of a single collaboration scheme becomes quite time-consuming, which makes it infeasible to evaluate all schemes and find an optimal scheme through traversal algorithm or directed acyclic graph optimization like the early works(Kang et al., 2017; Hu et al., 2019).”
A Comparative Analysis of Proactive and Reactivemethods for Privacy-Aware Interleaved Dnnoffloading
Cotter, J., Castiñeiras, I., O’Shea, D., & Cionca, V. (2022) · Computer Networks · doi:10.1016/j.comnet.2023.109999 · Cited by 1
DNN inference splitting and offloading in the Internet of Things : A survey
Bouarouguene, C., Maimour, M., Kadri, O., Rondeau, É., & Benyahia, A. (2023) · HAL · Cited by 1
He, X., Liu, Z., Kou, Z., Dong, N., Li, Y., & Liu, Z. (2023) · IEEE International Conference on Parallel and Distributed Systems (ICPADS) · doi:10.1109/icpads60453.2023.00150 · Cited by 1
“• DADS [7]: This approach employs the minimum cut algorithm on a modified computation graph to identify the optimal partitioning.”
MethodCoCoI: Distributed Coded Inference System for Straggler Mitigation
Liu, X., Huang, C. H., & Tang, M. (2025) · International Symposium on Modeling and Optimization in Mobile, Ad Hoc, and Wireless Networks (WiOpt) · doi:10.23919/wiopt66569.2025.11123371 · Cited by 1
“Some of them considered cloud-edge collaboration [6], [7], reducing inference workload on edge by assigning part of the model to the cloud.”
CloudEye: A New Paradigm of Video Analysis System for Mobile Visual ScenariosPreprint
Cui, H. Y., Li, Q., Wang, H., & jiang, Y. (2024) · arXiv (Cornell University) · doi:10.48550/arxiv.2410.18399 · Cited by 1
“In this approach, edge servers perform inference during cloud server inference gaps, but the weaker edge servers yield lower accuracy for inferred frames [15], and transmitting frames to cloud servers consumes considerable bandwidth [49, 50].”
BackgroundCheng, Z., Xia, X., Wang, H., Liwang, M., Chen, N., Fan, X., & Wang, X. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2502.16091 · Cited by 2
“For example, foundational researches such as [10], [21], [22] have established methodologies for model partition and inference optimization.”
CoCoI: Distributed Coded Inference System for Straggler MitigationPreprint
Liu, X., Huang, C., & Tang, M. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2501.06856 · Cited by 1
“Some of them considered cloud-edge collaboration [6], [7], reducing inference workload on edge by assigning part of the model to the cloud.”
Joint Optimization of Offloading, Batching and DVFS for Multiuser Co-InferencePreprint
Xu, Y., Zhou, S., & Niu, Z. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2504.14611 · Cited by 1
“However, existing studies on edge-device co-inference [3]–[5] cannot be directly applied to these DNN hardware accelerators.”
Distributed computing at the smart device edge
Feltin, T. (2023) · theses.fr (ABES) · doi:10.70675/2f7e1eb6zdb29z4aa3zb2d2za399e1df734b · Cited by 1
Lightweight Offloading System For Edge Cloud Environments
정혁진 (2020) · Seoul National University Open Repository (Seoul National University)
An adaptive DNN Partition Algorithm Greedy Algorithm (ADPA)
Zhanglei (2021) · IEEE International Conference on Software Engineering and Service Science (ICSESS) · doi:10.1109/icsess52187.2021.9522172
Building Secure and Reliable Deep Learning Systems from a Systems Security Perspective
Hong, S. (2021) · University Libraries (University of Maryland) · doi:10.13016/e1kn-sgys
Machine Learning-based Optimal Framework for Internet of Things Networks
Alsafasfeh, M. H., Arida, Z. A., & Saraereh, O. A. (2022) · Computers, materials & continua/Computers, materials & continua (Print) · doi:10.32604/cmc.2022.024093
“The authors in [23] designed an adaptive DNN splitting algorithm, which can find the optimal splitting strategy under dynamic and time-varying network load conditions.”
JointCS: Joint Search for Deep Model Compression and Segmentation on Heterogeneous IoT Devices
Li, X., Guo, B., Liu, S., Qiu, C., Liang, Y., & Yu, Z. (2021) · IEEE International Conference on Parallel and Distributed Systems (ICPADS) · doi:10.1109/icpads53394.2021.00059
Toward cooperative multi-agent video streaming perception
杨, 铮., 董, 亮., & 蔡, 新. (2023) · Scientia Sinica Informationis · doi:10.1360/ssi-2021-0179
“视频分析任务的构成是有向无环图 (direct acyclic graph, DAG) [36,37] , 相关 工作 [38,39] 提出将模型推断过程拆分成本地和云端分别运行的两部分 , 并降低两者之间的数据传输延 迟 .”
Deep learning inference time guarantee for aggregated images at the edge of the network
Genda, K. (2023) · IEICE Communications Express · doi:10.1587/comex.2023xbl0036
DeepSlicing: Collaborative and Adaptive CNN Inference With Low Latency
Zhang, S., Chen, Y., Zhang, S., & Chen, Z. (2023) · Advances in computational intelligence and robotics book series · doi:10.4018/979-8-3693-0230-9.ch006
Efficient Technique Implementation of Smart IoT Systems in Cloud Computing and Edge Used Systems
Ranganathaswamy, M. K., Vishnoi, M., & Khare, M. D. (2023) · International Conference on Power Energy, Environment and Intelligent Control (PEEIC) · doi:10.1109/peeic59336.2023.10451884
“[3] The limited battery life of IoT devices must be taken into consideration when making decisions about oflloading, and DNN-based smart loT system deadlines must be balanced with energy consumption on servers (both cloud and edge) and IoT devices.”
BackgroundMalathi, H., Malviya, A. V., & Umrao, S. (2024) · International Conference on Innovative Sustainable Technologies for Energy, Mechatronics, and Smart Systems (ISTEMS) · doi:10.1109/istems60181.2024.10560298
“This combination of state-of-the-art technologies offers a viable way to achieve ideal resource distributions in network designs of the future [39].”
BackgroundXu, J., Yang, Z., Liu, Y., & Cao, H. (2024) · Springer eBooks · doi:10.1007/978-981-97-3573-0_8
Application-Oriented Industrial Edge Computing
Zhou, X., Ge, S., Chi, J., & Qiu, T. (2024) · Springer eBooks · doi:10.1007/978-981-97-4752-8_6
Non-Linear Coded Computation for Distributed CNN Inference: A Learning-based Approach
Wang, H., Liu, Y., Xu, C., Liu, X., & Tang, M. (2024) · IEEE International Conference on Distributed Computing Systems Workshops (ICDCSW) · doi:10.1109/icdcsw63686.2024.00027
“As in many existing works on distributed inference (e.g. [12], [13], [14]), we consider the setting where F Conv is performed on n distributed devices, and F Fc is performed by an arbitrary device in a centralized fashion.”
BackgroundCorrections to “DNN Surgery: Accelerating DNN Inference on the Edge through Layer Partitioning”
Liang, H., Sang, Q., Hu, C., Cheng, D., Zhou, X., Wang, D., Bao, W., & Wang, Y. (2024) · IEEE Transactions on Cloud Computing · doi:10.1109/tcc.2024.3404548
Self-citationBovee, N., Piccolo, S., Bitla, S., Patapanchala, G. K., & Ho, S.-S. (2024) · IEEE International Conference on Fog and Edge Computing (ICFEC) · doi:10.1109/icfec61590.2024.00024
“These types of models better resemble directed acyclic graphs (DAG) as stated by Hu et. al [9], To accomplish our non-invasive approach, our initial probing of the DNN creates a map of which layers require input from multiple previous layers, and whether future layers will also require some…”
BackgroundOn the complexity of minimizing energy consumption of partitioning DAG tasks
Liu, W., Chen, J.-J., & Yang, Y. (2024) · Theoretical Computer Science · doi:10.1016/j.tcs.2024.115047
Collaborative Inference for Adaptive DNN Pipeline-Aware Alignment
Ren, J., Liu, L., Bi, R., Li, S., Li, S., & Lv, X. (2025) · Communications in computer and information science · doi:10.1007/978-981-96-0055-7_25
Reinforcement Learning based Edge-Assisted Inference for Maritime UAV Networks
Wang, C., Li, J., Ye, L., Liu, Y., & Xiao, L. (2024) · IEEE International Symposium on Parallel and Distributed Processing with Applications (ISPA) · doi:10.1109/ispa63168.2024.00222
“To reduce the processing delay for one frame under various DNN partition strategy, the DNN surgery scheme in [10] proposes a model partition solution to cut the raw DNN into several sets solely containing the source side and the sink side, reducing the latency and enhancing the throughput.”
Gemini+: Enhancing Real-Time Video Analytics With Dual-Image FPGAs
Wu, J., Hu, C., Zhang, J., Wang, D., & Jiang, J. (2025) · IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems · doi:10.1109/tcad.2025.3558464
“Neurosurgeon [35], DNN Surgery [36], and JALAB [37] partition DNN models between devices and servers, while FilterForward [38] accelerates video analytics by running lightweight NN models at the edge and reusing intermediate results.”
Self-citationRACE: Operator Choreography for Inference Acceleration in Personalized Recommender System
Li, S., Yang, X., Liu, J. H., Qi, Q., Liao, J., Huang, J., & Wang, J. (2025) · ACM Symposium on Parallelism in Algorithms and Architectures (SPAA) · doi:10.1145/3694906.3743310
DOPart: Competitive-Ratio Optimal Online Offloading of Sequentially Dependent Tasks
Saxena, S., & Liang, B. (2025) · IEEE International Conference on Machine Learning for Communication and Networking (ICMLCN) · doi:10.1109/icmlcn64995.2025.11140175
“We remark that the method in [10] behaves the same way as Neurosurgeon for sequentially dependent tasks, so this benchmark also represents [10].”
ObCLIP: Oblivious CLoud-Device Hybrid Image Generation with Privacy Preservation
Wu, H., Dai, W., Xu, M., Li, W., & Yan, Q. (2025) · neural information processing systems · doi:10.52202/085713-1097
AODPart: Accuracy-Optimal Online Partitioning for Edge Inference with Delay Constraint
Saxena, S., & Liang, B. (2025) · ACM International Symposium on Theory, Algorithmic Foundations, and Protocol Design for Mobile Networks and Mobile Computing (MobiHoc) · doi:10.1145/3704413.3764468
Parecon: Enhancing Mobile AI Vision Apps through DNN Partition and Resolution Control
Park, J., & Kim, Y. (2025) · IEEE International Conference on Mobile Ad-Hoc and Sensor Systems (MASS) · doi:10.1109/mass66014.2025.00067
“(iii) DADS : It chooses a partition point to maximize processed fps under heavy network loads or to minimize E2E latency under light network loads [14].”
OSLCNN: On-Demand Scheduling for Large-Scale CNN Inference Acceleration on Heterogeneous DevicesPreprint
Yang, L., zhong, D., Liu, S., Zheng, C., Li, R., & 程, 鑫. (2025) · SSRN Electronic Journal · doi:10.2139/ssrn.5859191
Design-Space Exploration of Distributed Neural Networks in Low-Power Wearable NodesPreprint
Chowdhury, M. R., Li, M., Ghosh, A., Bari, M. F., & Sen, S. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2509.14540
Tang, S., Rui, H., Luo, H., Wang, C., Yu, C., Li, Y., Fu, H., Sun, C., & Xiao, A. J. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2507.15628
Li, Y. (2023) · ANU Open Research (Australian National University) · doi:10.25911/kfnt-0p96
System support & orchestration mechanisms for distributed DNN inference
Reisinger, M. (2022) · reposiTUm (TU Wien) · doi:10.34726/hss.2022.87400
Bringing Agile and Self-Evolvable Intelligence to Weak Embedded Devices
Huang, K. (2024) · D-Scholarship@Pitt (University of Pittsburgh)
Machine Learning Model Splitting on Mobile Edge Networks
Wang, S. (2023) · eScholarship (California Digital Library)
PEPPER: Profiling-based Edge Placement and Partitioning for Deep Learning Execution
Korontanis, I., Kontopoulos, I., Zacharia, A., Makris, A., Chronis, C., Pateraki, M., Tserpes, K., & Varlamis, I. (2025) · International Conference on the Internet of Things (IoT) · doi:10.1145/3770501.3770528
“The authors compare their approach against contemporary model splitting approaches, namely the Neurosurgeon approach described above, and DADS [16], a graph-theoretic approach similar to QDMP.”
Accelerating Object Detection and Tracking Pipelines for Efficient Edge Video Analytics
Xu, R. (2025) · MacSphere (McMaster University) · doi:10.71548/16615
Adaptive Collaborative DNN Inference via Heterogeneous Graph Attention in Edge Environments
Shi, C., Jiang, B., Xu, H., Xie, J., & Wang, C. (2025) · IEEE Internet of Things Journal · doi:10.1109/jiot.2025.3648047
“Consequently, early collaborative inference approaches such as Neurosurgeon [4], DeepThings [5], and DADS [6], designed for simple, chain-structured models in static settings, have become insu ffi cient.”
Priority-Aware DNN Offloading via Queuing Latency Estimation in Multi-User Edge-Device System
Tao, G., Li, X., & Qin, X. (2025) · Conference proceedings/Conference proceedings - IEEE International Conference on Systems, Man, and Cybernetics · doi:10.1109/smc58881.2025.11342891
A Security Analysis of CNN Partitioning Strategies for Distributed Inference at the Edge
Mehrafrooz, F., Siyadatzadeh, R., Mentens, N., & Stefanov, T. P. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-16342-4_15
“Another approach is to distribute a CNN model by running a part of it on an edge device and the rest on a cloud server [6].”
Adaptive DNN Partitioning Strategy for Optimized User Fitness in Edge Computing Networks
Zheng, X., Tahir, M., Tang, Y., Anwar, M. S., Hussain, D. I., & Choi, A. (2026) · IEEE Transactions on Consumer Electronics · doi:10.1109/tce.2026.3663558
“In DNN models, the output data volume of some intermediate layers is much smaller than the original input data volume, so this method has been widely used in DNN model inference [8].”
Reinforcement Learning Based Model Adaptation and Partition for Dynamic DNN Empowered Mobile Vision
Liu, Y., Xiao, L., Li, T., Lv, Z., Qiao, Z., & Xiong, H. (2026) · IEEE Transactions on Mobile Computing · doi:10.1109/tmc.2026.3665503
“Despite chain-structure model splitting, the modern DNN with more complex structure is formulated as a DAG and split the model via an approximation-based algo-rithm [19].”
JMPS: A Joint Model Partitioning and Scheduling Approach for DNN Inference in Mobile Edge Networks
Fang, J., Tang, H. L., Teng, Z., Shi, Z., Wu, S., & An, Y. (2025) · International Conference on Computing, Communication, Perception and Quantum Technology (CCPQT) · doi:10.1109/ccpqt66408.2025.11383368
“In order to fully simulate the communication between end devices and servers in mobile edge environments, we refer to the experimental setup of [10] and test the average upload rate under several different wireless networks (e.g., CAT1, 3G, 4G, 5G and WIFI), as shown in Table I. DNN benchmarks.”
InfluentialEfficient Deployment of AI Applications on Edge Devices
Hu, C. (2025) · TSpace (University of Toronto)
Self-citationKayal, P. (2024) · TSpace (University of Toronto)
Edge and Cloud-Based AI Inference for Big Data Processing
Elmobark, N. (2026) · Advances in computational intelligence and robotics book series · doi:10.4018/979-8-3373-7046-0.ch004
Hybrid Partitioning System for Embedded and Distributed CNNs Inference on Edge Devices.
Kaboubi, N. (2023) · theses.fr (ABES) · doi:10.70675/b1e891ddz8b9cz4c57zb475z9bf0adff96d4
Labriji, I. (2022) · theses.fr (ABES) · doi:10.70675/679a9526z8a9cz4c36za387z6e7571424f32
Distributed Optimization and Machine Learning for Virtualized 6G Wireless Networks
Ehsanian, A. (2024) · theses.fr (ABES) · doi:10.70675/bb4d9bcfzf764z4fc6z9189z86b3a2717498
Adaptive Microservice Deployment for Data Partition-Based Applications in MEC
Huang, Z., Dong, F., Ruan, W., Guo, X., & Zhang, J. (2026) · Tsinghua Science & Technology · doi:10.26599/tst.2024.9010255
End-Edge-Cloud Orchestration for Video Analytics
Bai, T. (2025) · Textbooks in telecommunication engineering · doi:10.1007/978-3-032-14764-6_6
Delta Coding With Dead-Zone Quantization for Bandwidth-Efficient Split Computing on Video Streams
Shiba, H., Zhu, Z., Finnerty, P., & Ohta, C. (2026) · IEEE Access · doi:10.1109/access.2026.3690477
“DADS [12] introduces dynamic DNN surgery that adjusts the split point at runtime based on network conditions and device load.”
Net4LMs: Wireless Network-Aided Language Models
Du, H., Chen, X., Liu, Y., & Huang, K. (2026) · Wireless networks · doi:10.1007/978-3-032-20194-2_4
Zhang, X., Zhang, L., Liu, M., Ge, M., & Ling, L. (2026) · Journal of Industrial Information Integration · doi:10.1016/j.jii.2026.101134
Poster: Accelerating Offloaded ViT Inference via Progressive Refinement
Park, C., Park, G., Joe‐Wong, C., & Lee, K. (2026) · Annual International Conference on Mobile Systems, Applications, and Services Companion (MobiSys Companion) · doi:10.1145/3812835.3814856
“Unlike prior split computing [2, 3] and pipelined offloading systems [1, 5], which primarily exploit CNN-oriented structures, progressive refinement explores a new pipelining dimension through spectral decomposition of the input.”
ZHAO, Y., Mo, Z., He, Q., Niu, Z., Cao, J., & Hawbani, A. (2026) · IEEE Transactions on Computers · doi:10.1109/tc.2026.3699819
“Research in DNN model partitioning is evolving from handling simple chain-structured models to addressing models with complex directed acyclic graph (DAG) topologies [22].”
CANS: Accelerating Multiuser Collaborative Edge Inference via Cooperative Autodidactic NeuroSurgeonPreprint
Wu, Z., Zhang, Z., Lin, C., Xu, Z., & Liu, J. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2606.09175
Reliable collaborative inference for embodied AI empowered by application-aware IPv6 networking
Chen, J., Fu, Z., Shang, Q., & Ma, H. (2026) · Journal of Information and Intelligence · doi:10.1016/j.jiixd.2026.06.004
Facility-Location-based Model Partitioning for Efficient Inference in Edge Intelligence
Ma, W., Ren, C., & Ai, C. (2026) · Global Information Infrastructure and Networking Symposium · doi:10.1109/giis69881.2026.11585754
“(2) Current segmentation strategies employ static partitioning schemes based solely on model characteristics without adapting to dynamic network conditions [16]–[18].”
Multi-Turn Distributed Inference with Mixture of Experts for 6G Edge--Cloud NetworksPreprint
Liu, B., Li, H., Liu, Y., Wu, Y., Hussain, R., Moazzeni, S., & Simeonidou, D. E. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2607.02522
“DADS [5] formulates partitioning as a constrained optimization and adapts the split point as bandwidth and load conditions evolve.”
Device Assignment and Model Splitting Optimization for Resilient and Secure Multi-Hop Split Learning
Wei, D., Zhou, D., Liu, Y., & Chen, M. (2026) · IEEE Transactions on Wireless Communications · doi:10.1109/twc.2026.3713859
“However, most of these works [5], [6], [7], [8], [9], [10], [11], [12] did not consider the risk of training information being monitored by eavesdroppers during multi-hop transmissions.”
InfluentialQoS Parameters in Edge Computing Environment for Internet of Things Applications - An Overview
Bhise, S., Kshirsagar, K., & Deshpande, V. (2026) · Lecture notes in networks and systems · doi:10.1007/978-3-032-23716-3_18
Myroshnyk, Y., & Biriukov, Y. (2026) · IEEE Access · doi:10.1109/access.2026.3719609
“Neurosurgeon [8] and Dynamic Adaptive DNN Surgery (DADS) [12] partition deep neural network (DNN) layers between a mobile device and an edge server, choosing the split point to minimize latency given the network bandwidth.”
Tian, Y., Jiang, T., Tian, K., Zhou, R., Liu, Y., Zhou, Z., Dong, F., Liu, M., & Liu, W. (2026) · IEEE International Conference on Distributed Computing Systems (ICDCS) · doi:10.1109/2575-8411.2026.00135
In-Browser Split Inference for Autoregressive Large Language Models With Activation Quantization
Lee, Y., Choee, Y.-S., Lee, S.-H., & Kwon, S. (2026) · IEEE Access · doi:10.1109/access.2026.3720562
“Although this approach has been widely explored for convolutional neural networks (CNNs) in edge computing [2], [3], applying it to in-browser LLM scenarios introduces distinct challenges.”
CST: Collaborative Selective Transmission for Communication-Efficient Multimodal Edge InferencePreprint
Chi, H., Zhang, J., Ning, R., Wang, C., Gazda, R., Yang, H., & Wu, H. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.22115
“Split computing and pipelined inference optimize single-model execution across device–edge/cloud resources [7], [35].”
Practical Modeling for Split DNN Inference on Near-Edge Accelerators
Liu, H., Fouda, M. E., Eltawil, A. M., & Fahmy, S. A. (2026) · IEEE International Conference on Edge Computing and Communications (EDGE) · doi:10.1109/edge72783.2026.00011
“Existing split DNN search space optimization algorithms [3], [18]–[20] all rely on such fundamental computation and communication models.”
TAILOR: Token-Aware Partitioning and Routing for Edge–Cloud Transformer Inference
Huang, X., Liu, R. R., & Wu, J. (2026) · International Workshop on Quality of Service · doi:10.1109/iwqos70441.2026.11661040
“Early systems such as Neurosur-geon [7], DADS [16], and Auto-Split [6] optimize split points or offloading decisions under heterogeneous computation and network conditions.”
Fine-Grained DNN Partitioning and Offloading with Cold Start Mitigation in Vehicular Edge Computing
Nie, Y., Zhang, Y., Jin, H., & Zhang, P. (2026) · IEEE International Conference on Software Services Engineering (SSE) · doi:10.1109/sse72781.2026.00024
“Existing DNN offloading methods often rely on shortest-path or heuristic strategies, leading to coarse partitions and limited adaptability to dynamic links and heterogeneous resources [6] [7] [8].”
Adaptive Scheduling with Dynamic Heterogeneous Graphs for Intelligent Cloud-Edge Computing Services
Chen, J., & Wang, J. (2026) · IEEE International Conference on Web Services (ICWS) · doi:10.1109/icws72778.2026.00176
“DADS [5] extends partitioning from chain structured DNNs to graph structured models, enabling collaborative inference for more complex dependency structures.”
Deployment-Aware Codec Selection for Learned RGB-D Compression in Edge--Cloud 3D ReconstructionPreprint
Zhang, Y., Zhang, Z., & Zhang, Z. (2026) · Research Square · doi:10.21203/rs.3.rs-10347179/v1
Costa, D., & Lima, R. (2025) · International Journal of Modern Computer Science and IT Innovations
DS-Net: A Dynamic Split Network for Edge-Cloud Collaborative InferencePreprint
Zhang, X., Wang, L., Jin, L., Xu, L., & Han, Z. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.23949
“Other approaches [18]–[21] have adopted graph-based modeling techniques.”
From Partitioning to Provisioning: Co-Designing DNN Deployment and Infrastructure CapacityPreprint
Huang, Z., Yang, F., Cao, Q., & Shao, Y. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.27749 · Cited by 1
“The slowest stage determines the output interval, whose reciprocal gives the inference throughput [8].”
Workflow-Aware Communication and Computation Scheduling for Mobile Edge AI AgentsPreprint
He, X., You, C., Wang, F., Chen, Z., Wang, S., & Park, J. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.29337
“DADS further adapts DNN partitioning between edge and cloud resources to changing network conditions [18].”
TandemServe: Network-adaptive prefill orchestration for edge–cloud LLM serving
Park, S. S., Kim, D., Park, J., & Oh, Y. H. (2026) · Journal of Systems Architecture · doi:10.1016/j.sysarc.2026.104064
STLT: A Novel System for Energy-Aware Distributed Inference in the Edge-Cloud-Space ContinuumPreprint
Stanisic, A., Gravara, M., Herrera, J. L., & Nastić, S. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2611.01233
AMSDS: Adaptive Cloud-Edge Visual Navigation via DNN Partitioning and Semantic Compression
Hao, Y., Ding, N., Sun, W., & Xu, L. (2026) · ACM Transactions on Embedded Computing Systems · doi:10.1145/3860325
Metadata
- Authors
- Hu, C., Bao, W., Wang, D., & Liu, F.
- Affiliations
- Hong Kong Polytechnic University (HK)
- The University of Sydney (AU)
- Venue
- IEEE INFOCOM 2019 - IEEE Conference on Computer Communications
- Publisher
- IEEE
- Published
- 2019-04
- Pages
- 1423-1431
- OpenAlex
- W2920031528
- Cited by
- 416
- References
- 23
- Citations by year
- 2026: 29 · 2025: 61 · 2024: 65 · 2023: 92 · 2022: 62 · 2021: 72 · 2020: 26 · 2019: 9
- FWCI
- 36.52
- Type
- conference paper
- Language
- en
- Open access
- closed
- Licence
- Locations
- IEEE Conference on Computer Communications (INFOCOM) (publishedVersion)
- Rare & Special e-Zone (The Hong Kong University of Science and Technology) (submittedVersion)
- Topics
- Physical Sciences › Computer Science › Computer Networks and Communications › IoT and Edge/Fog Computing
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
- Physical Sciences › Computer Science › Hardware and Architecture › Cloud Computing and Remote Desktop Technologies
- Keywords
- DNN partitioning, cloud-edge collaboration, inference acceleration, dynamic network conditions, directed acyclic graphs, minimum cut, approximation algorithms, throughput maximization, latency reduction, data transmission reduction
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.