mlsanityData to 7 Oct 2026

Ekya: Continuous Learning of Video Analytics Models on Edge Compute Servers

arXiv 2012.10557 ↗First posted 19 Dec 2020cs.DC cs.AIcited by 58

Not on the map. Why?

Abstract (arXiv)

Video analytics applications use edge compute servers for the analytics of the videos (for bandwidth and privacy). Compressed models that are deployed on the edge servers for inference suffer from data drift, where the live video data diverges from the training data. Continuous learning handles data drift by periodically retraining the models on new data. Our work addresses the challenge of jointly supporting inference and retraining tasks on edge servers, which requires navigating the fundamental tradeoff between the retrained model's accuracy and the inference accuracy. Our solution Ekya balances this tradeoff across multiple models and uses a micro-profiler to identify the models that will benefit the most by retraining. Ekya's accuracy gain compared to a baseline scheduler is 29% higher, and the baseline requires 4x more GPU resources to achieve the same accuracy as Ekya.

Key facts. Cited by 58 papers in the map (data through 7 Oct 2026). First posted on arXiv 19 Dec 2020.

Builds on

No reference list yet.

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2021: 2; 2022: 8; 2023: 13; 2024: 16; 2025: 16; 2026: 3.

58 papers.

Sort:

By area:

  1. 2024Cost-Effective Online Multi-LLM Selection with Versatile Reward Modelscited by 36
  2. 2022Federated Learning under Distributed Concept Driftcited by 35
  3. 2022GEMEL: Model Merging for Memory-Efficient, Real-Time Video Analytics at the Edgecited by 32
  4. 2024Galaxy: A Resource-Efficient Collaborative Edge AI System for In-situ Transformer Inferencecited by 25
  5. 2023EdgeFM: Leveraging Foundation Model for Open-set Learning on the Edgecited by 21
  6. 2022On-device Training: A First Overview on Existing Systemscited by 18
  7. 2023Apparate: Rethinking Early Exits to Tame Latency-Throughput Tensions in ML Servingcited by 15
  8. 2024Asteroid: Resource-Efficient Hybrid Pipeline Parallelism for Collaborative DNN Training on Heterogeneous Edge Devicescited by 13
  9. 2022Spatial-Temporal Federated Learning for Lifelong Person Re-identification on Distributed Edgescited by 12
  10. 2023Shoggoth: Towards Efficient Edge-Cloud Collaborative Real-Time Video Inference via Adaptive Online Learningcited by 11
  11. 2021Boggart: Towards General-Purpose Acceleration of Retrospective Video Analyticscited by 10
  12. 2023Cost-effective On-device Continual Learning over Memory Hierarchy with Mirocited by 10
38 more
  1. 2024Continual Learning for Smart City: A Surveycited by 9
  2. 2023Enabling Resource-efficient AIoT System with Cross-level Optimization: A surveycited by 8
  3. 2024DaCapo: Accelerating Continuous Learning in Autonomous Systems for Video Analyticscited by 8
  4. 2022Turbo: Opportunistic Enhancement for Edge Video Analyticscited by 7
  5. 2022Edge Video Analytics: A Survey on Applications, Systems and Enabling Techniquescited by 7
  6. 2023MadEye: Boosting Live Video Analytics Accuracy with Adaptive Camera Configurationscited by 7
  7. 2023Monitoring and Adapting ML Models on Mobile Devicescited by 6
  8. 2024Delta: A Cloud-assisted Data Enrichment Framework for On-Device Continual Learningcited by 6
  9. 2021Chanakya: Learning Runtime Decisions for Adaptive Real-Time Perceptioncited by 5
  10. 2023Modyn: Data-Centric Machine Learning Pipeline Orchestrationcited by 5
  11. 2025Compass: SLO-aware Query Planner for Compound AI Serving at Scalecited by 5
  12. 2025AVA: Towards Agentic Video Analytics with Vision Language Modelscited by 5
  13. 2023View while Moving: Efficient Video Recognition in Long-untrimmed Videoscited by 4
  14. 2024AxiomVision: Accuracy-Guaranteed Adaptive Visual Model Selection for Perspective-Aware Video Analyticscited by 4
  15. 2023Moby: Empowering 2D Models for Efficient Point Cloud Analytics on the Edgecited by 3
  16. 2024A Scene-aware Models Adaptation Scheme for Cross-scene Online Inference on Mobile Devicescited by 3
  17. 2024Federated Learning Clients Clustering with Adaptation to Data Driftscited by 3
  18. 2023OneAdapt: Fast Configuration Adaptation for Video Analytics Applications via Backpropagationcited by 2
  19. 2024Empower Vision Applications with LoRA LMMcited by 2
  20. 2025A Survey on Video Analytics in Cloud-Edge-Terminal Collaborative Systemscited by 2
  21. 2025E4: Energy-Efficient DNN Inference for Edge Video Analytics Via Early-Exit and DVFScited by 2
  22. 2025LeMix: Unified Scheduling for LLM Training and Inference on Multi-GPU Systemscited by 2
  23. 2023EdgeMA: Model Adaptation System for Real-Time Video Analytics on Edge Devicescited by 1
  24. 2023ECLM: Efficient Edge-Cloud Collaborative Learning with Continuous Environment Adaptationcited by 1
  25. 2024Enabling Cross-Camera Collaboration for Video Analytics on Distributed Smart Camerascited by 1
  26. 2024EdgeOL: Efficient in-situ Online Learning on Edge Devicescited by 1
  27. 2024On Distributed Larger-Than-Memory Subset Selection With Pairwise Submodular Functionscited by 1
  28. 2024EdgeSync: Faster Edge-model Updating via Adaptive Continuous Learning for Video Data Driftcited by 1
  29. 2025Prism: Mining Task-aware Domains in Non-i.i.d. IMU Data for Flexible User Perceptioncited by 1
  30. 2025GPUs, CPUs, and... NICs: Rethinking the Network's Role in Serving Complex AI Pipelinescited by 1
  31. 2025High-Quality Pseudo-Label Generation Based on Visual Prompt Assisted Cloud Model Updatecited by 1
  32. 2025MACE: A Hybrid LLM Serving System with Colocated SLO-aware Continuous Retraining Alignmentcited by 1
  33. 2025Cerberus: Real-Time Video Anomaly Detection via Cascaded Vision-Language Modelscited by 1
  34. 2022Variant Parallelism: Lightweight Deep Convolutional Models for Distributed Inference on IoT Devicescited by 0
  35. 2022APT: Adaptive Perceptual quality based camera Tuning using reinforcement learningcited by 0
  36. 2024Compass: A Decentralized Scheduler for Latency-Sensitive ML Workflowscited by 0
  37. 2024Online Resource Allocation for Edge Intelligence with Colocated Model Retraining and Inferencecited by 0
  38. 2025Alchemist: Towards the Design of Efficient Online Continual Learning Systemcited by 0

50 most cited of 58 All 58 (JSON)