LEO: scheduling sensor inference algorithms across heterogeneous mobile processors and network resources
The DOI matches its Crossref record; 48 of the 48 references it cites were found in a registry or an index.
Abstract
Mobile apps that use sensors to monitor user behavior often employ resource heavy inference algorithms that make computational offloading a common practice. However, existing schedulers/offloaders typically emphasize one primary offloading aspect without fully exploring complementary goals (e.g., heterogeneous resource management with only partial visibility into underlying algorithms, or concurrent sensor app execution on a single resource) and as a result, may overlook performance benefits pertinent to sensor processing.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Citations
Cites
Hinton, G., Deng, L., Yu, D., Dahl, G., Mohamed, A.-R., Jaitly, N., Senior, A., Vanhoucke, V., Nguyen, P., Sainath, T., et al. (2012) · IEEE Signal Processing Magazine · doi:10.1109/msp.2012.2205597 · Cited by 10,260
Pattern Recognition and Machine Learning
(2006) · doi:10.1007/978-0-387-45528-0 · Cited by 9,824
(2006) · Springer eBooks · Cited by 8,123
Perceptual linear predictive (PLP) analysis of speech
Hermansky, H. (1990) · The Journal of the Acoustical Society of America · doi:10.1121/1.399423 · Cited by 2,536 · cited 3×
“One example of shared features are the Perceptual Linear Predictive Coefficients (PLP) needed by both Speaker Iden-tification and Emotion Recognition applications [61].”
BackgroundA survey of mobile phone sensing
Lane, N., Miluzzo, E., Lu, H., Peebles, D., Choudhury, T., & Campbell, A. (2010) · IEEE Communications Magazine · doi:10.1109/mcom.2010.5560598 · Cited by 2,292 · cited 1×
“A rapidly growing trend among smartphone apps is to enhance functionality and provide advanced features in near real-time using inferences from the phone’s sensor data [46].”
BackgroundSelf-citationMAUI: 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
Deep Neural Networks for Acoustic Modeling in Speech Recognition
Hinton, G. E., Deng, L., Yu, D., Dahl, G. E., Mohamed, A., Jaitly, N., Senior, A., Vanhoucke, V., Nguyen, P., Sainath, T. N., et al. (2012) · IEEE Signal Processing Magazine · Cited by 2,136
YIN, a fundamental frequency estimator for speech and music
de Cheveigné, A., & Kawahara, H. (2002) · The Journal of the Acoustical Society of America · doi:10.1121/1.1458024 · Cited by 1,920 · cited 1×
“and Time Domain J48 Decision Tree 4s 4s Step Counting [28] Accel step counting Time Domain Windowed Peak Thresholding 4s 4s Speaker Count [66] Mic speaker counting MFCC [35], pitch [33] Unsupervised Clustering 32ms 3s Emotion Rec.”
Method(2011) · doi:10.1145/1966445.1966473 · Cited by 1,870
On Evolution, Search, Optimization, Genetic Algorithms and Martial Arts : Towards Memetic Algorithms
Moscato, P. (1989) · Cited by 1,516
Studies in Computational Intelligence
Ortíz-Arroyo, D. (2014) · Studies in computational intelligence · Cited by 1,193
Comparison of different implementations of MFCC
Zheng, F., Zhang, G., & Song, Z. (2001) · Journal of Computer Science and Technology · doi:10.1007/bf02943243 · Cited by 591 · cited 1×
“and Time Domain J48 Decision Tree 4s 4s Step Counting [28] Accel step counting Time Domain Windowed Peak Thresholding 4s 4s Speaker Count [66] Mic speaker counting MFCC [35], pitch [33] Unsupervised Clustering 32ms 3s Emotion Rec.”
MethodThe Jigsaw continuous sensing engine for mobile phone applications
Lu, H., Yang, J., Liu, Z., Lane, N. D., Choudhury, T., & Campbell, A. T. (2010) · Proceedings of the 8th ACM Conference on Embedded Networked Sensor Systems · doi:10.1145/1869983.1869992 · Cited by 560
Self-citationSmall-footprint keyword spotting using deep neural networks
Chen, G., Parada, C., & Heigold, G. (2014) · 2014 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) · doi:10.1109/icassp.2014.6854370 · Cited by 545 · cited 2×
“Second, the scheduler employs a fast heuristic optimization solver (based on metaheuristics) that is executed with an ultra low overhead on the LPU to find a near optimal assignment of tasks to resources.”
MethodMyLifeBits: a personal database for everything
Gemmell, J., Bell, G., & Lueder, R. (2006) · Communications of the ACM · doi:10.1145/1107458.1107460 · Cited by 516 · cited 1×
“Most apps of this type are “life-logging” [36] and are commonly used to quantify daily user routines.”
BackgroundStressSense: detecting stress in unconstrained acoustic environments using smartphones
Lu, H., Frauendorfer, D., Rabbi, M., Mast, M. S., Chittaranjan, G. T., Campbell, A. T., Gatica-Perez, D., & Choudhury, T. (2012) · Proceedings of the 2012 ACM Conference on Ubiquitous Computing · doi:10.1145/2370216.2370270 · Cited by 511 · cited 2×
“However, such cases are expected to be rare since mobile sensor processing [51, 61, 50, 53, 66, 52] is typically periodic over the sensor stream with short repeated tasks to maintain tight mobile resource consumption and timeliness guarantees.”
BackgroundEmotionSense: a mobile phones based adaptive platform for experimental social psychology research
Rachuri, K. K., Musolesi, M., Mascolo, C., Rentfrow, P. J., Longworth, C., & Aucinas, A. (2010) · Proceedings of the 12th ACM international conference on Ubiquitous computing · doi:10.1145/1864349.1864393 · Cited by 508 · cited 4×
“For instance, the inference stage of a typical Speaker Identification application [61] is usually organized around multiple class models (Gaussian Mixture Models [27]) as shown in Figure 2a, one for each speaker.”
BackgroundMethodInfluentialSelf-citationDeepX: A Software Accelerator for Low-Power Deep Learning Inference on Mobile Devices
Lane, N. D., Bhattacharya, S., Georgiev, P., Forlivesi, C., Jiao, L., Qendro, L., & Kawsar, F. (2016) · 2016 15th ACM/IEEE International Conference on Information Processing in Sensor Networks (IPSN) · doi:10.1109/ipsn.2016.7460664 · Cited by 450
Self-citationMoodScope: building a mood sensor from smartphone usage patterns
LiKamWa, R., Liu, Y., Lane, N. D., & Zhong, L. (2013) · Proceeding of the 11th annual international conference on Mobile systems, applications, and services · doi:10.1145/2462456.2464449 · Cited by 444 · cited 1×
“…Mic conversation context SpeakerSense [50] Crowd++ [66], SocialWeaver [53] EmotionSense [61] Mic emotion recognition StressSense [51], MoodScope [48] Siri [6], Cortana [13] Mic digital assistants Waze [26], Moovit [15] GPS traffic monitoring Just like conventional apps, different combinations of…”
BackgroundSelf-citationNonlinear feature based classification of speech under stress.
Zhou, G., Hansen, J. H. L., & Kaiser, J. F. (2001) · IEEE Transactions on Speech and Audio Processing · doi:10.1109/89.905995 · Cited by 409
Aurasium: practical policy enforcement for Android applications
Xu, R., Saïdi, H., & Anderson, R. (2012) · Cited by 378
Walk detection and step counting on unconstrained smartphones
Brajdic, A., & Harle, R. (2013) · Proceedings of the 2013 ACM international joint conference on Pervasive and ubiquitous computing · doi:10.1145/2493432.2493449 · Cited by 375 · cited 1×
“and Time Domain J48 Decision Tree 4s 4s Step Counting [28] Accel step counting Time Domain Windowed Peak Thresholding 4s 4s Speaker Count [66] Mic speaker counting MFCC [35], pitch [33] Unsupervised Clustering 32ms 3s Emotion Rec.”
MethodOdessa: enabling interactive perception applications on mobile devices
Ra, M.-R., Sheth, A., Mummert, L., Pillai, P., Wetherall, D., & Govindan, R. (2011) · Proceedings of the 9th international conference on Mobile systems, applications, and services · doi:10.1145/1999995.2000000 · Cited by 361 · cited 2×
“General-purpose offload-ers [32, 59, 62, 69, 30, 60, 38], do not target the diverse sensor processing workloads explicitly and, as a result, may miss substantial optimization opportunities.”
BackgroundAnatomizing application performance differences on smartphones
Huang, J., Xu, Q., Tiwana, B., Mao, Z. M., Zhang, M., & Bahl, P. (2010) · Proceedings of the 8th international conference on Mobile systems, applications, and services · doi:10.1145/1814433.1814452 · Cited by 296 · cited 1×
“4 Mbps [41]).”
Rachuri, K. K., Mascolo, C., Musolesi, M., & Rentfrow, P. J. (2011) · Proceedings of the 17th annual international conference on Mobile computing and networking · doi:10.1145/2030613.2030623 · Cited by 197 · cited 2×
“General-purpose offload-ers [32, 59, 62, 69, 30, 60, 38], do not target the diverse sensor processing workloads explicitly and, as a result, may miss substantial optimization opportunities.”
BackgroundSelf-citationKang, S., Lee, J., Jang, H., Lee, H., Lee, Y., Park, S., Park, T., & Song, J. (2008) · Proceedings of the 6th international conference on Mobile systems, applications, and services · doi:10.1145/1378600.1378630 · Cited by 185 · cited 1×
“Existing sensor orchestra-tor frameworks [44, 42, 43, 49, 52, 45] approach the optimization space from different angles in order to improve sensor processing on resource-constraint mobile devices, and often provide complementary functionality that can be executed side by side with LEO.”
BackgroundAppJoy: personalized mobile application discovery
Yan, B., & Chen, G. (2011) · Proceedings of the 9th international conference on Mobile systems, applications, and services · doi:10.1145/1999995.2000007 · Cited by 171
LittleRock: Enabling Energy-Efficient Continuous Sensing on Mobile Phones
Priyantha, B., Lymberopoulos, D., & Liu, J. (2011) · IEEE Pervasive Computing · doi:10.1109/mprv.2011.28 · Cited by 161 · cited 1×
“SpeakerSense [50], Little Rock [58], DSP.Ear [37] and Audio-DAQ [64] utilize low-power co-processors or purpose-built peripheral devices to achieve energy savings while supporting a fixed set of constantly running sensing applications.”
BackgroundDevice Analyzer: Understanding Smartphone Usage
Wagner, D. T., Rice, A. C., & Beresford, A. R. (2014) · Lecture notes of the Institute for Computer Sciences, Social Informatics and Telecommunications Engineering · doi:10.1007/978-3-319-11569-6_16 · Cited by 158 · cited 2×
“For example, a survey of more than 12 , 000 devices worldwide [65] finds that a sizable 20 % of the devices are not exposed to 3G, LTE or WiFi connectivity at least 45 % of the time.”
BackgroundSpeakerSense: Energy Efficient Unobtrusive Speaker Identification on Mobile Phones
Lu, H., Brush, A. J. B., Priyantha, N. B., Karlson, A. K., & Liu, J. (2011) · Lecture notes in computer science · doi:10.1007/978-3-642-21726-5_12 · Cited by 155 · cited 1×
“Every sensor app includes specialized code responsible for sensor data sampling and processing, dis-Applications”
MethodJust-in-time provisioning for cyber foraging
Ha, K., Pillai, P., Richter, W., Abe, Y., & Satyanarayanan, M. (2013) · Proceeding of the 11th annual international conference on Mobile systems, applications, and services · doi:10.1145/2462456.2464451 · Cited by 146 · cited 2×
“General-purpose offload-ers [32, 59, 62, 69, 30, 60, 38], do not target the diverse sensor processing workloads explicitly and, as a result, may miss substantial optimization opportunities.”
BackgroundCrowd++: unsupervised speaker count with smartphones
Xu, C., Li, S., Liu, G., Zhang, Y., Miluzzo, E., Chen, Y.-F., Li, J., & Firner, B. (2013) · Proceedings of the 2013 ACM international joint conference on Pervasive and ubiquitous computing · doi:10.1145/2493432.2493435 · Cited by 122 · cited 2×
“However, such cases are expected to be rare since mobile sensor processing [51, 61, 50, 53, 66, 52] is typically periodic over the sensor stream with short repeated tasks to maintain tight mobile resource consumption and timeliness guarantees.”
BackgroundWishbone: profile-based partitioning for sensornet applications
Newton, R., Toledo, S., Girod, L., Balakrishnan, H., & Madden, S. R. (2009) · DSpace@MIT (Massachusetts Institute of Technology) · Cited by 109
Kang, S., Lee, Y., Min, C., Ju, Y., Park, T., Lee, J., Rhee, Y., & Song, J. (2010) · IEEE International Conference on Pervasive Computing and Communications · doi:10.1109/percom.2010.5466982 · Cited by 79 · cited 3×
“Orchestrator [44] does not scale well with the increase in number of offloading configurations as it systematically explores sub-sets of offloading plans the number of which grows exponentially with offloading components and sensor apps.”
BackgroundRavindranath, L., Thiagarajan, A., Balakrishnan, H., & Madden, S. R. (2012) · IEEE Workshop on Mobile Computing Systems and Applications (WMCSA) · doi:10.1145/2162081.2162087 · Cited by 65 · cited 2×
“General-purpose offload-ers [32, 59, 62, 69, 30, 60, 38], do not target the diverse sensor processing workloads explicitly and, as a result, may miss substantial optimization opportunities.”
BackgroundDSP.Ear: leveraging co-processor support for continuous audio sensing on smartphones
Georgiev, P., Lane, N. D., Rachuri, K. K., & Mascolo, C. (2014) · Proceedings of the 12th ACM Conference on Embedded Network Sensor Systems · doi:10.1145/2668332.2668349 · Cited by 60
Self-citationReflex: using low-power processors in smartphones without knowing them
Lin, F. X., Wang, Z., LiKamWa, R., & Zhong, L. (2012) · International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS) · doi:10.1145/2150976.2150979 · Cited by 60 · cited 2×
“Reflex [49], for example, eases programmability for LPUs [49] but does not explicitly optimize for energy efficiency.”
BackgroundNirjon, S., Dickerson, R. F., Asare, P., Li, Q., Hong, D., Stankovic, J. A., Hu, P., Shen, G., & Jiang, X. (2013) · International Conference on Mobile Systems, Applications, and Services · doi:10.1145/2462456.2464446 · Cited by 51 · cited 1×
“Conventional offloading applied to mobile sensing (e.g., [57]) must rely on the CPU for local operations.”
CAreDroid: Adaptation Framework for Android Context-Aware Applications
Elmalaki, S., Wanner, L., & Srivastava, M. B. (2015) · ACM Annual International Conference on Mobile Computing and Networking (MobiCom) · doi:10.1145/2789168.2790108 · Cited by 48 · cited 1×
“CAreDroid [34], on the other hand, presents a framework for automatically choosing among different implementations of the same sensor processing logic that leads to highest performance gains given the current device and user context.”
BackgroundCustomizable and extensible deployment for mobile/cloud applications
Zhang, I. Y., Szekeres, A., Aken, D. V., Ackerman, I., Gribble, S. D., Krishnamurthy, A., & Levy, H. M. (2014) · Operating Systems Design and Implementation · doi:10.5555/2685048.2685057 · Cited by 46
SocialWeaver: collaborative inference of human conversation networks using smartphones
Luo, C., & Chan, M. C. (2013) · ACM Conference on Embedded Networked Sensor Systems (SenSys) · doi:10.1145/2517351.2517353 · Cited by 45 · cited 3×
“…Mic voice activated services Shazam [24] Mic song recognition SocioPhone [47] Mic conversation context SpeakerSense [50] Crowd++ [66], SocialWeaver [53] EmotionSense [61] Mic emotion recognition StressSense [51], MoodScope [48] Siri [6], Cortana [13] Mic digital assistants Waze [26], Moovit [15]…”
BackgroundSymPhoney: a coordinated sensing flow execution engine for concurrent mobile sensing applications
Ju, Y., Lee, Y., Yu, J., Min, C., Shin, I., & Song, J. (2012) · ACM Conference on Embedded Networked Sensor Systems (SenSys) · doi:10.1145/2426656.2426678 · Cited by 35 · cited 5×
“Existing sensor orchestra-tor frameworks [44, 42, 43, 49, 52, 45] approach the optimization space from different angles in order to improve sensor processing on resource-constraint mobile devices, and often provide complementary functionality that can be executed side by side with LEO.”
BackgroundInfluentialMemetic Algorithms in Planning, Scheduling, and Timetabling
Cotta, C., & Fernández, A. J. (2007) · Studies in computational intelligence · doi:10.1007/978-3-540-48584-1_1 · Cited by 34 · cited 2×
“The concrete framework we use is based on memetic algorithms (MAs) [55, 31] which are population-based metaheuristics that are highly scalable, are able to incorporate do-main-specific knowledge, and are less prone to local optima.”
MethodEnhancing mobile apps to use sensor hubs without programmer effort
Shen, H., Balasubramanian, A., LaMarca, A., & Wetherall, D. J. (2015) · ACM International Joint Conference on Pervasive and Ubiquitous Computing (UbiComp) · doi:10.1145/2750858.2804260 · Cited by 30 · cited 3×
“A variety of capable scheduling/offloading approaches have been proposed [56, 32, 54, 42, 63] but they either have different optimization goals or have not fully addressed the above question in the emerging context of concurrent sensor apps with diverse deadlines running on recent off-the-shelf…”
BackgroundMethodKrishnan, S. P. T., & Gonzalez, J. L. U. (2015) · Apress eBooks · doi:10.1007/978-1-4842-1004-8_11 · Cited by 27
Verma, S., Robinson, A. L., & Dutta, P. (2012) · ACM Conference on Embedded Networked Sensor Systems (SenSys) · doi:10.1145/2426656.2426677 · Cited by 22 · cited 1×
“SpeakerSense [50], Little Rock [58], DSP.Ear [37] and Audio-DAQ [64] utilize low-power co-processors or purpose-built peripheral devices to achieve energy savings while supporting a fixed set of constantly running sensing applications.”
BackgroundSocioPhone: everyday face-to-face interaction monitoring platform using multi-phone sensor fusion
Lee, Y., Min, C., Hwang, C., Lee, J., Hwang, I., Ju, Y., Yoo, C., Moon, M., Lee, U., & Song, J. (2013) · International Conference on Mobile Systems, Applications, and Services · doi:10.1145/2462456.2465702 · Cited by 14
ORBIT: a smartphone-based platform for data-intensive embedded sensing applications
Moazzami, M.-M., Phillips, D. E., Tan, R., & Xing, G. (2015) · ACM/IEEE International Conference on Information Processing in Sensor Networks (IPSN) · doi:10.1145/2737095.2737098 · Cited by 13 · cited 4×
“A variety of capable scheduling/offloading approaches have been proposed [56, 32, 54, 42, 63] but they either have different optimization goals or have not fully addressed the above question in the emerging context of concurrent sensor apps with diverse deadlines running on recent off-the-shelf…”
BackgroundMethodInfluential
Research works printed but not linked: 2
- 10.5555/2959355.2959378 · doi:10.5555/2959355.2959378
- 10.5555/2021975.2021992 · doi:10.5555/2021975.2021992
Cited by
Edge Intelligence: Paving the Last Mile of Artificial Intelligence With Edge Computing
Zhou, Z., Chen, X., Li, E., Zeng, L., Luo, K., & Zhang, J. (2019) · Proceedings of the IEEE · doi:10.1109/jproc.2019.2918951 · Cited by 2,455
“LEO [113] is a novel sensing algorithm scheduler that maximizes the performance for multiple continuous mobile sensor applications by partitioning the sensing algorithm execution and distributing tasks on CPU, coprocessor, GPU and the cloud.”
BackgroundEdge 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
“Model Problem Object Latency Energy consumption [305] DNN Model partition Acceleration, save energy 23× 85.5% reduction [306] DNN Open source framework Enable edge inference N/A N/A [307] AlexNet,VGGNet Incremental training Improve accuracy 1.4− 3.3× 30%-70% saving [308] DNN Task scheduling Minimise latency 1.2− 1.7×…”
BackgroundInfluentialOn-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
DeepWear: Adaptive Local Offloading for On-Wearable Deep Learning
Xu, M., Qian, F., Zhu, M., Huang, F., Pushp, S., & Liu, X. (2020) · IEEE Transactions on Mobile Computing · doi:10.1109/tmc.2019.2893250 · Cited by 138
“M AKING deep learning (DL for short in the rest of this paper) tasks run on mobile devices has raised huge interests in both the academia [1], [2], [3], [4], [5], [6], [7], [8] and the industry [9], [10], [11].”
Background
Metadata
- Authors
- Georgiev, P., Lane, N. D., Rachuri, K. K., & Mascolo, C.
- Affiliations
- University of Cambridge (GB)
- University College London (GB)
- Samsung (United States) (US)
- University of Cambridge
- ORCID
- Lane, Nicholas D.
- Venue
- Proceedings of the 22nd Annual International Conference on Mobile Computing and Networking
- Publisher
- ACM
- Published
- 2016-10-03
- Pages
- 320-333
- ISSN
- 1543-5679 (ISSN-L)
- OpenAlex
- W2525851376
- Cited by
- 58
- References
- 47
- Citations by year
- 2025: 1 · 2024: 4 · 2023: 3 · 2022: 4 · 2021: 9 · 2020: 9 · 2019: 10 · 2018: 13 · 2017: 5
- FWCI
- 7.97
- Type
- conference paper
- Language
- en
- Open access
- closed
- Locations
- ACM Annual International Conference on Mobile Computing and Networking (MobiCom) (publishedVersion)
- Topics
- Physical Sciences › Computer Science › Computer Networks and Communications › IoT and Edge/Fog Computing
- Physical Sciences › Computer Science › Information Systems › Mobile and Web Applications
- Physical Sciences › Computer Science › Hardware and Architecture › Parallel Computing and Optimization Techniques
- Keywords
- computation offloading, inference scheduling, heterogeneous resource management
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.