Neurosurgeon: Collaborative Intelligence Between the Cloud and Mobile Edge

By Yiping Kang, Johann Hauswald, Cao Gao, A. Rovinski, T. Mudge, Jason Mars, Lingjia Tang. 2017

1,544 citations recorded by the source corpus.

Subfields: Computer Networks and Communications, Computer Vision and Pattern Recognition, Efficiency & Compression

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Networks and Communications

In Computer Vision and Pattern Recognition

In Efficiency & Compression

Citations in the corpus

Sustainable AI: Environmental Implications, Challenges and Opportunities (2021, background): “Optimizing the overall energy efficiency of FL and on-device AI is an important first step [89], [90], [91], [92], [93].”
Chasing Carbon: The Elusive Environmental Footprint of Computing (2020, background): “Others have proposed novel scheduler designs to enable energy-efﬁcient, collaborative cloud and edge execution [69], [92].”
Edge Intelligence: Paving the Last Mile of Artificial Intelligence With Edge Computing (2019, background): “For the model partition between the server and the device, Neurosurgeon [79] represents an iconic effort.”
Communication-Efficient Edge AI: Algorithms and Systems (2020, methodology): “In addition, some computation-intensive AI inference tasks can only be accomplished by task-splitting among edge devices and edge servers [16], which also incurs heavy communication cost.”
Efficient Acceleration of Deep Learning Inference on Resource-Constrained Edge Devices: A Review (2023, methodology): “In sequential distribution, DL layers are assigned to different edge devices [25], [374].”
Privacy-Preserving Inference in Machine Learning Services Using Trusted Execution Environments (2019, background): “Neurosurgeon [19] partitions DNNs and onloads them to mobile devices to reduce latency and energy consumption.”
Incentive Mechanism Design for Resource Sharing in Collaborative Edge Learning (2020, neutral): “Note that the DNN partitioning can be optimized in consideration of the energy constraints of user devices [9].”
Overlearning Reveals Sensitive Attributes (2019, background): “The motivations are better scalability of inference [16, 20] and protecting privacy of inputs into the model [3, 23, 30, 35].”
DeepThings: Distributed Adaptive Deep Learning Inference on Resource-Constrained IoT Edge Clusters (2018, methodology): “In [9] and [10], DNN inference is partially offloaded to cloud servers to minimize processing latency and edge device resource consumption.”
AutoScale: Energy Efficiency Optimization for Stochastic Edge Inference Using Reinforcement Learning (2020, methodology): “AutoScale signiﬁcantly improves energy efﬁciency compared with the baselines and prior work [42],[53] satisfying QoS constraints. action.”
CoEdge: Cooperative DNN Inference With Adaptive Workload Partitioning Over Heterogeneous Edge Devices (2020, background): “Neurosurgeon [42] proposes a partially offloading solution, which decides an intermediate partition point in the DNN structure to keep front layers locally and offload rear layers to the cloud.”
Convergence of Edge Computing and Deep Learning: A Comprehensive Survey (2019, methodology): “For a computation-intensive task, it will be reasonably segmented and dispatched separately to the end, edge and cloud for execution, reducing the execution delay of the task while ensuring the accuracy of the results [12], [49], [50].”
Understanding Capacity-Driven Scale-Out Neural Recommendation Inference (2020, background): “Cooperative, hybrid approaches–that split the model be-tween mobile systems and data-center–have also been explored for improved performance and efﬁciency [42].”
Orbital Edge Computing: Nanosatellite Constellations as a New Class of Computer System (2020, background): “video querying [44], search [57], or DNN speech processing [55].”
AutoScale: Optimizing Energy Efficiency of End-to-End Edge Inference under Stochastic Variance (2020, methodology): “This layer is one of the most compute- and memory-intensive layers in DNNs [15,44,46], since its neurons are exhaustively connected to all the neurons in the previous layer.”
A Survey of Wearable Devices and Challenges (2017, background): “The approach can be further extended as highlighted in a recent research [279] where authors proposed a system called Neurosurgeon that dynamically partitions the neural network”