Pub Date : 2020-11-01DOI: 10.1109/SEC50012.2020.00022
Alireza Mohtadi, Julien Gascon-Samson
In the last few years, the number of IoT applications that rely on stream processing has increased significantly. These applications process continuous streams of data with a low delay and provide valuable information. To meet the stringent latency requirements and the need for real-time results that they require, the components of the stream processing pipeline can be deployed directly onto the edge layer to benefit from the resources and capabilities that the swarm of edge devices can provide. In this poster, we outline some ongoing research ideas into deploying stream processing operators onto edge nodes, with the goal of minimizing latency while ensuring that the constraints of the devices and their network capabilities are respected. More precisely, we provide a modeling of the semantics of the operators that considers the interactions between different operators, the parallelism of concurrent operators, as well as the latency and bandwidth usage.
{"title":"Poster: Dependency-Aware Operator Placement of Distributed Stream Processing IoT Applications Deployed at the Edge","authors":"Alireza Mohtadi, Julien Gascon-Samson","doi":"10.1109/SEC50012.2020.00022","DOIUrl":"https://doi.org/10.1109/SEC50012.2020.00022","url":null,"abstract":"In the last few years, the number of IoT applications that rely on stream processing has increased significantly. These applications process continuous streams of data with a low delay and provide valuable information. To meet the stringent latency requirements and the need for real-time results that they require, the components of the stream processing pipeline can be deployed directly onto the edge layer to benefit from the resources and capabilities that the swarm of edge devices can provide. In this poster, we outline some ongoing research ideas into deploying stream processing operators onto edge nodes, with the goal of minimizing latency while ensuring that the constraints of the devices and their network capabilities are respected. More precisely, we provide a modeling of the semantics of the operators that considers the interactions between different operators, the parallelism of concurrent operators, as well as the latency and bandwidth usage.","PeriodicalId":375577,"journal":{"name":"2020 IEEE/ACM Symposium on Edge Computing (SEC)","volume":"47 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117114271","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-11-01DOI: 10.1109/SEC50012.2020.00051
Li-Tse Hsieh, Hang Liu, Yang Guo, Robert Gazda
This paper investigates the task management for cooperative mobile edge computing (MEC), where a set of geographically distributed heterogeneous edge nodes not only cooperate with remote cloud data centers but also help each other to jointly process tasks and support real-time IoT applications at the edge of the network. Especially, we address the challenges in optimizing assignment of the tasks to the nodes under dynamic network environments when the task arrivals, node computing capabilities, and network states are nonstationary and unknown a priori. We propose a novel stochastic framework to model the interactions of the involved entities, including the edge-to-edge horizontal cooperation and the edge-to-cloud vertical cooperation. The task assignment problem is formulated and the algorithm is developed based on online reinforcement learning to optimize the performance for task processing while capturing various dynamics and heterogeneities of node computing capabilities and network conditions with no requirement for prior knowledge of them. Further, by leveraging the structure of the underlying problem, a post-decision state is introduced and a function decomposition technique is proposed, which are incorporated with reinforcement learning to reduce the search space and computation complexity. The evaluation results demonstrate that the proposed online learning-based scheme outperforms the state-of-the-art benchmark algorithms.
{"title":"Task Management for Cooperative Mobile Edge Computing","authors":"Li-Tse Hsieh, Hang Liu, Yang Guo, Robert Gazda","doi":"10.1109/SEC50012.2020.00051","DOIUrl":"https://doi.org/10.1109/SEC50012.2020.00051","url":null,"abstract":"This paper investigates the task management for cooperative mobile edge computing (MEC), where a set of geographically distributed heterogeneous edge nodes not only cooperate with remote cloud data centers but also help each other to jointly process tasks and support real-time IoT applications at the edge of the network. Especially, we address the challenges in optimizing assignment of the tasks to the nodes under dynamic network environments when the task arrivals, node computing capabilities, and network states are nonstationary and unknown a priori. We propose a novel stochastic framework to model the interactions of the involved entities, including the edge-to-edge horizontal cooperation and the edge-to-cloud vertical cooperation. The task assignment problem is formulated and the algorithm is developed based on online reinforcement learning to optimize the performance for task processing while capturing various dynamics and heterogeneities of node computing capabilities and network conditions with no requirement for prior knowledge of them. Further, by leveraging the structure of the underlying problem, a post-decision state is introduced and a function decomposition technique is proposed, which are incorporated with reinforcement learning to reduce the search space and computation complexity. The evaluation results demonstrate that the proposed online learning-based scheme outperforms the state-of-the-art benchmark algorithms.","PeriodicalId":375577,"journal":{"name":"2020 IEEE/ACM Symposium on Edge Computing (SEC)","volume":"29 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123612929","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-11-01DOI: 10.1109/SEC50012.2020.00021
Joel Wolfrath, A. Chandra
Data stream processing is an increasingly important topic due to the prevalence of smart devices and the demand for realtime analytics. One estimate suggests that we should expect nine smart-devices per person by the year 2025 [1]. These devices generate data which might include sensor readings from a smart home, event or system logs on a device, or video feeds from surveillance cameras. As the number of devices increases, the cost of streaming the device data to the cloud over the wide-area network (WAN) will also increase substantially. Transferring and querying this data efficiently has become the focus of much academic research [2]–[5]. Edge computation affords us the opportunity to address this problem by utilizing resources close to the devices. Edge resources have many different use cases, including minimizing end-to-end latency or maximizing throughput [6], [7]. We restrict our focus to minimizing the required WAN bandwidth, which is an effort to address the increase in data volume.
{"title":"Poster: Data-Aware Edge Sampling for Aggregate Query Approximation","authors":"Joel Wolfrath, A. Chandra","doi":"10.1109/SEC50012.2020.00021","DOIUrl":"https://doi.org/10.1109/SEC50012.2020.00021","url":null,"abstract":"Data stream processing is an increasingly important topic due to the prevalence of smart devices and the demand for realtime analytics. One estimate suggests that we should expect nine smart-devices per person by the year 2025 [1]. These devices generate data which might include sensor readings from a smart home, event or system logs on a device, or video feeds from surveillance cameras. As the number of devices increases, the cost of streaming the device data to the cloud over the wide-area network (WAN) will also increase substantially. Transferring and querying this data efficiently has become the focus of much academic research [2]–[5]. Edge computation affords us the opportunity to address this problem by utilizing resources close to the devices. Edge resources have many different use cases, including minimizing end-to-end latency or maximizing throughput [6], [7]. We restrict our focus to minimizing the required WAN bandwidth, which is an effort to address the increase in data volume.","PeriodicalId":375577,"journal":{"name":"2020 IEEE/ACM Symposium on Edge Computing (SEC)","volume":"16 9","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132531300","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Continuing the symposium's tradition, this year we have strived to put together an exciting program of high quality papers, keynote talks and panel session. The programme is structured around the general themes of Distributed Simulation, Virtual Reality and Virtual and Telepresent Humans, and Military simulation with sessions comprising excellent papers that truly represent state-of-the-art research, technologies and applications. These papers were selected after hard and long evaluation process. This year accepted invited papers came from:
{"title":"Message from the Program Co-Chairs","authors":"G. Theodoropoulos","doi":"10.1109/ds-rt.2011.5","DOIUrl":"https://doi.org/10.1109/ds-rt.2011.5","url":null,"abstract":"Continuing the symposium's tradition, this year we have strived to put together an exciting program of high quality papers, keynote talks and panel session. The programme is structured around the general themes of Distributed Simulation, Virtual Reality and Virtual and Telepresent Humans, and Military simulation with sessions comprising excellent papers that truly represent state-of-the-art research, technologies and applications. These papers were selected after hard and long evaluation process. This year accepted invited papers came from:","PeriodicalId":375577,"journal":{"name":"2020 IEEE/ACM Symposium on Edge Computing (SEC)","volume":"13 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115737740","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
With the rapid growth of IoT devices, the traditional cloud computing scheme is inefficient for many IoT based applications, mainly due to network data flood, long latency, and privacy issues. To this end, the edge computing scheme is proposed to mitigate these problems. However, in an edge computing system, the application development becomes more complicated as it involves increasing levels of edge nodes. Although some efforts have been introduced, existing edge computing frameworks still have some limitations in various application scenarios. To overcome these limitations, we propose a new programming model called Edge-Stream. It is a simple and programmer-friendly model, which can cover typical scenarios in edge-computing. Besides, we address several new issues, such as data sharing and area awareness, in this model. We also implement a prototype of edge-computing framework based on the Edge-Stream model. A comprehensive evaluation is provided based on the prototype. Experimental results demonstrate the effectiveness of the model.
{"title":"Edge-Stream: a Stream Processing Approach for Distributed Applications on a Hierarchical Edge-computing System","authors":"Xiaoyang Wang, Zhe Zhou, Ping Han, Tong Meng, Guangyu Sun, Jidong Zhai","doi":"10.1109/SEC50012.2020.00009","DOIUrl":"https://doi.org/10.1109/SEC50012.2020.00009","url":null,"abstract":"With the rapid growth of IoT devices, the traditional cloud computing scheme is inefficient for many IoT based applications, mainly due to network data flood, long latency, and privacy issues. To this end, the edge computing scheme is proposed to mitigate these problems. However, in an edge computing system, the application development becomes more complicated as it involves increasing levels of edge nodes. Although some efforts have been introduced, existing edge computing frameworks still have some limitations in various application scenarios. To overcome these limitations, we propose a new programming model called Edge-Stream. It is a simple and programmer-friendly model, which can cover typical scenarios in edge-computing. Besides, we address several new issues, such as data sharing and area awareness, in this model. We also implement a prototype of edge-computing framework based on the Edge-Stream model. A comprehensive evaluation is provided based on the prototype. Experimental results demonstrate the effectiveness of the model.","PeriodicalId":375577,"journal":{"name":"2020 IEEE/ACM Symposium on Edge Computing (SEC)","volume":"41 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123958646","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-11-01DOI: 10.1109/SEC50012.2020.00050
Winson Ye, Qun Li
Over the last few years, there has been a growing interest in developing chatbots that can converse intelligently with humans. For example, consider Microsoft’s Xiaoice. It is a highly intelligent dialogue system that serves as both a social companion and a virtual assistant. Targeted towards Chinese users, Xiaoice is connected to 660 million online users and 450 million IoT devices. Because of the deep learning revolution, the field is moving quickly, so this survey aims to introduce newcomers to the most fundamental research questions for next generation neural dialogue systems. In particular, our analysis of the state of the art reveals the following 4 key research challenges: 1) knowledge grounding, 2) persona consistency, 3) emotional intelligence, and 4) evaluation. Knowledge grounding endows the chatbot with external knowledge to generate more informative replies. Persona consistency grants dialogue systems consistent personalities. We divide each fundamental research challenge into several smaller and more concrete research questions. For each fine grained research challenge, we examine state of the art approaches and propose future research directions.
{"title":"Open Questions for Next Generation Chatbots","authors":"Winson Ye, Qun Li","doi":"10.1109/SEC50012.2020.00050","DOIUrl":"https://doi.org/10.1109/SEC50012.2020.00050","url":null,"abstract":"Over the last few years, there has been a growing interest in developing chatbots that can converse intelligently with humans. For example, consider Microsoft’s Xiaoice. It is a highly intelligent dialogue system that serves as both a social companion and a virtual assistant. Targeted towards Chinese users, Xiaoice is connected to 660 million online users and 450 million IoT devices. Because of the deep learning revolution, the field is moving quickly, so this survey aims to introduce newcomers to the most fundamental research questions for next generation neural dialogue systems. In particular, our analysis of the state of the art reveals the following 4 key research challenges: 1) knowledge grounding, 2) persona consistency, 3) emotional intelligence, and 4) evaluation. Knowledge grounding endows the chatbot with external knowledge to generate more informative replies. Persona consistency grants dialogue systems consistent personalities. We divide each fundamental research challenge into several smaller and more concrete research questions. For each fine grained research challenge, we examine state of the art approaches and propose future research directions.","PeriodicalId":375577,"journal":{"name":"2020 IEEE/ACM Symposium on Edge Computing (SEC)","volume":"18 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124211458","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-11-01DOI: 10.1109/SEC50012.2020.00067
Bodong Shang, Shiya Liu, Sidi Lu, Y. Yi, Weisong Shi, Lingjia Liu
In Internet-of-Thing (IoT) networks, enormous low-power IoT devices execute latency-sensitive yet computationintensive machine learning tasks. However, the energy is usually scarce for IoT devices, especially for some without battery and relying on solar power or other renewables forms. In this paper, we introduce a cross-layer optimization framework for distributed computing among low-power IoT devices. Specifically, a programming layer design for distributed IoT networks is presented by addressing the problems of application partition, task scheduling, and communication overhead mitigation. Furthermore, the associated federated learning and local differential privacy schemes are developed in the communication layer to enable distributed machine learning with privacy preservation. In addition, we illustrate a three-dimensional network architecture with various network components to facilitate efficient and reliable information exchange among IoT devices. Moreover, a model quantization design for IoT devices is illustrated to reduce the cost of information exchange. Finally, a parallel and scalable neuromorphic computing system for IoT devices is established to achieve energy-efficient distributed computing platforms in the hardware layer. Based on the introduced cross-layer optimization framework, IoT devices can execute their machine learning tasks in an energy-efficient way while guaranteeing data privacy and reducing communication costs.
{"title":"A Cross-Layer Optimization Framework for Distributed Computing in IoT Networks","authors":"Bodong Shang, Shiya Liu, Sidi Lu, Y. Yi, Weisong Shi, Lingjia Liu","doi":"10.1109/SEC50012.2020.00067","DOIUrl":"https://doi.org/10.1109/SEC50012.2020.00067","url":null,"abstract":"In Internet-of-Thing (IoT) networks, enormous low-power IoT devices execute latency-sensitive yet computationintensive machine learning tasks. However, the energy is usually scarce for IoT devices, especially for some without battery and relying on solar power or other renewables forms. In this paper, we introduce a cross-layer optimization framework for distributed computing among low-power IoT devices. Specifically, a programming layer design for distributed IoT networks is presented by addressing the problems of application partition, task scheduling, and communication overhead mitigation. Furthermore, the associated federated learning and local differential privacy schemes are developed in the communication layer to enable distributed machine learning with privacy preservation. In addition, we illustrate a three-dimensional network architecture with various network components to facilitate efficient and reliable information exchange among IoT devices. Moreover, a model quantization design for IoT devices is illustrated to reduce the cost of information exchange. Finally, a parallel and scalable neuromorphic computing system for IoT devices is established to achieve energy-efficient distributed computing platforms in the hardware layer. Based on the introduced cross-layer optimization framework, IoT devices can execute their machine learning tasks in an energy-efficient way while guaranteeing data privacy and reducing communication costs.","PeriodicalId":375577,"journal":{"name":"2020 IEEE/ACM Symposium on Edge Computing (SEC)","volume":"362 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122767802","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-11-01DOI: 10.1109/SEC50012.2020.00025
Bingqian Lu, Jianyi Yang, Shaolei Ren
Deep neural networks (DNNs) have been increasingly deployed on and integrated with edge devices, such as mobile phones, drones, robots and wearables. Compared to cloud-based inference, running DNN inference directly on edge devices (a.k. a. edge inference) has major advantages, including being free from the network connection requirement, saving bandwidths, and better protecting user privacy [1].
{"title":"Poster: Scaling Up Deep Neural Network optimization for Edge Inference†","authors":"Bingqian Lu, Jianyi Yang, Shaolei Ren","doi":"10.1109/SEC50012.2020.00025","DOIUrl":"https://doi.org/10.1109/SEC50012.2020.00025","url":null,"abstract":"Deep neural networks (DNNs) have been increasingly deployed on and integrated with edge devices, such as mobile phones, drones, robots and wearables. Compared to cloud-based inference, running DNN inference directly on edge devices (a.k. a. edge inference) has major advantages, including being free from the network connection requirement, saving bandwidths, and better protecting user privacy [1].","PeriodicalId":375577,"journal":{"name":"2020 IEEE/ACM Symposium on Edge Computing (SEC)","volume":"4 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122556240","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-11-01DOI: 10.1109/SEC50012.2020.00015
M. Chao, R. Stoleru, Liuyi Jin, Shuochao Yao, Maxwell Maurice, R. Blalock
The popularity of video cameras has spawned a new type of application called multitask video processing, which uses multiple CNNs to obtain different information of interests from a raw video stream. Unfortunately, the huge resource requirements of CNNs make the concurrent execution of multiple CNNs on a single resource-constrained mobile device challenging. Existing solutions solve this challenge by offloading CNN models to the cloud or edge server, compressing CNN models to fit the mobile device, or sharing some common parts of multiple CNN models. Most of these solutions, however, use the above offloading, compression or sharing strategies in a separate manner, which fail to adapt to the complex edge computing scenario well. In this paper, to solve the above limitation, we propose AMVP, an adaptive execution framework for CNN-based multitask video processing, which elegantly integrates the strategies of CNN layer sharing, feature compression, and model offloading. First, AMVP reduces the total computation workload of multiple CNN inference by sharing some common frozen CNN layers. Second, AMVP supports distributed CNN inference by splitting big CNNs into smaller components running on different devices. Third, AMVP leverages a quantization-based feature compression mechanism to reduce the feature transmission traffic size between two separate CNN components. We conduct extensive experiments on AMVP and the experimental results show that our AMVP framework can adapt to different performance goals and execution environments. Compared to two baseline approaches that only share or offload CNN layers, AMVP achieves up to 61% lower latency and 10% higher throughput with comparative accuracy.
{"title":"AMVP: Adaptive CNN-based Multitask Video Processing on Mobile Stream Processing Platforms","authors":"M. Chao, R. Stoleru, Liuyi Jin, Shuochao Yao, Maxwell Maurice, R. Blalock","doi":"10.1109/SEC50012.2020.00015","DOIUrl":"https://doi.org/10.1109/SEC50012.2020.00015","url":null,"abstract":"The popularity of video cameras has spawned a new type of application called multitask video processing, which uses multiple CNNs to obtain different information of interests from a raw video stream. Unfortunately, the huge resource requirements of CNNs make the concurrent execution of multiple CNNs on a single resource-constrained mobile device challenging. Existing solutions solve this challenge by offloading CNN models to the cloud or edge server, compressing CNN models to fit the mobile device, or sharing some common parts of multiple CNN models. Most of these solutions, however, use the above offloading, compression or sharing strategies in a separate manner, which fail to adapt to the complex edge computing scenario well. In this paper, to solve the above limitation, we propose AMVP, an adaptive execution framework for CNN-based multitask video processing, which elegantly integrates the strategies of CNN layer sharing, feature compression, and model offloading. First, AMVP reduces the total computation workload of multiple CNN inference by sharing some common frozen CNN layers. Second, AMVP supports distributed CNN inference by splitting big CNNs into smaller components running on different devices. Third, AMVP leverages a quantization-based feature compression mechanism to reduce the feature transmission traffic size between two separate CNN components. We conduct extensive experiments on AMVP and the experimental results show that our AMVP framework can adapt to different performance goals and execution environments. Compared to two baseline approaches that only share or offload CNN layers, AMVP achieves up to 61% lower latency and 10% higher throughput with comparative accuracy.","PeriodicalId":375577,"journal":{"name":"2020 IEEE/ACM Symposium on Edge Computing (SEC)","volume":"53 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124930193","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2020-11-01DOI: 10.1109/SEC50012.2020.00043
Fuxun Yu, Dimitrios Stamoulis, Di Wang, Dimitrios Lymberopoulos, Xiang Chen
This paper gives an overview of our ongoing work on the design space exploration of efficient deep neural networks (DNNs), specifically on the novel optimization perspectives that past work have mainly overlooked. We cover two complementary aspects of efficient DNN design: (1) static architecture design efficiency and (2) dynamic model execution efficiency. In the static architecture design, one of the major challenges of NAS is the low search efficiency. Different with current mainstream efficient search algorithm optimization, we identify the new perspective in efficient search space design. In the dynamic model execution, current major optimization methods still target at the model structure redundancy, e.g., weight/filter pruning, connection pruning, etc. We instead identify the new dimension of DNN feature map redundancy. By showcasing such new perspectives, further advantages could be potentially attained by integrating both current optimizations and our new perspectives.
{"title":"Exploring the Design Space of Efficient Deep Neural Networks","authors":"Fuxun Yu, Dimitrios Stamoulis, Di Wang, Dimitrios Lymberopoulos, Xiang Chen","doi":"10.1109/SEC50012.2020.00043","DOIUrl":"https://doi.org/10.1109/SEC50012.2020.00043","url":null,"abstract":"This paper gives an overview of our ongoing work on the design space exploration of efficient deep neural networks (DNNs), specifically on the novel optimization perspectives that past work have mainly overlooked. We cover two complementary aspects of efficient DNN design: (1) static architecture design efficiency and (2) dynamic model execution efficiency. In the static architecture design, one of the major challenges of NAS is the low search efficiency. Different with current mainstream efficient search algorithm optimization, we identify the new perspective in efficient search space design. In the dynamic model execution, current major optimization methods still target at the model structure redundancy, e.g., weight/filter pruning, connection pruning, etc. We instead identify the new dimension of DNN feature map redundancy. By showcasing such new perspectives, further advantages could be potentially attained by integrating both current optimizations and our new perspectives.","PeriodicalId":375577,"journal":{"name":"2020 IEEE/ACM Symposium on Edge Computing (SEC)","volume":"14 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125296064","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：