mlsanityData to 7 Oct 2026

It's always personal: Using Early Exits for Efficient On-Device CNN Personalisation

arXiv 2102.01393 ↗First posted 2 Feb 2021cs.LG cs.CVcited by 18

Not on the map. Why?

Abstract (arXiv)

On-device machine learning is becoming a reality thanks to the availability of powerful hardware and model compression techniques. Typically, these models are pretrained on large GPU clusters and have enough parameters to generalise across a wide variety of inputs. In this work, we observe that a much smaller, personalised model can be employed to fit a specific scenario, resulting in both higher accuracy and faster execution. Nevertheless, on-device training is extremely challenging, imposing excessive computational and memory requirements even for flagship smartphones. At the same time, on-device data availability might be limited and samples are most frequently unlabelled. To this end, we introduce PersEPhonEE, a framework that attaches early exits on the model and personalises them on-device. These allow the model to progressively bypass a larger part of the computation as more personalised data become available. Moreover, we introduce an efficient on-device algorithm that trains the early exits in a semi-supervised manner at a fraction of the whole network's personalisation time. Results show that PersEPhonEE boosts accuracy by up to 15.9% while dropping the training cost by up to 2.2x and inference latency by 2.2-3.2x on average for the same accuracy, depending on the availability of labels on-device.

Key facts. Cited by 18 papers in the map (data through 7 Oct 2026). Builds on 17 of its 28 references that are in the map. First posted on arXiv 2 Feb 2021.

Builds on

17 of 28 references.

References · 17

  1. 2015Deep Residual Learning for Image Recognitioncited by 51,809
  2. 2015Distilling the Knowledge in a Neural Networkcited by 10,076
  3. 2016Communication-Efficient Learning of Deep Networks from Decentralized Datacited by 7,592
  4. 2017On Calibration of Modern Neural Networkscited by 4,860
  5. 2018MobileNetV2: Inverted Residuals and Linear Bottleneckscited by 4,773
  6. 2020Meta-Learning in Neural Networks: A Surveycited by 944
  7. 2019Towards Federated Learning at Scale: System Designcited by 826
  8. 2017BranchyNet: Fast Inference via Early Exiting from Deep Neural Networkscited by 591
  9. 2019Towards Fairer Datasets: Filtering and Balancing the Distribution of the People Subtree in the ImageNet Hierarchycited by 156
  10. 2020SPINN: Synergistic Progressive Inference of Neural Networks over Device and Cloudcited by 109
  11. 2019AI Benchmark: All About Deep Learning on Smartphones in 2019cited by 76
  12. 2018Online Model Distillation for Efficient Video Inferencecited by 60
5 more
  1. 2020DarkneTZ: Towards Model Privacy at the Edge using Trusted Execution Environmentscited by 57
  2. 2020HAPI: Hardware-Aware Progressive Inferencecited by 30
  3. 2019SCAN: A Scalable Neural Networks Framework Towards Compact and Efficient Modelscited by 22
  4. 2019EmBench: Quantifying Performance Variations of Deep Neural Networks across Modern Commodity Devicescited by 21
  5. 2020Now that I can see, I can improve: Enabling data-driven finetuning of CNNs on the edgecited by 4

Built on by

Cited by year

Citing papers per year: 2021: 6; 2022: 5; 2023: 2; 2024: 2; 2025: 3.

18 papers.

Sort:

By area:

  1. 2021Adaptive Inference through Early-Exit Networks: Design, Challenges and Directionscited by 74
  2. 2021Smart at what cost? Characterising Mobile Deep Neural Networks in the wildcited by 25
  3. 2023EdgeFM: Leveraging Foundation Model for Open-set Learning on the Edgecited by 21
  4. 2021Multi-Exit Semantic Segmentation Networkscited by 18
  5. 2021On-device Federated Learning with Flowercited by 13
  6. 2022Benchmarking of DL Libraries and Models on Mobile Devicescited by 12
  7. 2024Effective Heterogeneous Federated Learning via Efficient Hypernetwork-based Weight Generationcited by 9
  8. 2025HELIOS: Adaptive Model And Early-Exit Selection for Efficient LLM Inference Servingcited by 8
  9. 2022T-RECX: Tiny-Resource Efficient Convolutional neural networks with early-eXitcited by 5
  10. 2022Federated Learning for Inference at Anytime and Anywherecited by 5
  11. 2022Mind Your Heart: Stealthy Backdoor Attack on Dynamic Deep Neural Network in Edge Computingcited by 5
  12. 2021perf4sight: A toolflow to model CNN training performance on Edge GPUscited by 4
6 more
  1. 2024DyCE: Dynamically Configurable Exiting for Deep Learning Compression and Real-time Scalingcited by 3
  2. 2025Energy-Aware Deep Learning on Resource-Constrained Hardwarecited by 3
  3. 2025MMEdge: Accelerating On-device Multimodal Inference via Pipelined Sensing and Encodingcited by 3
  4. 2021How to Reach Real-Time AI on Consumer Devices? Solutions for Programmable and Custom Architecturescited by 2
  5. 2022Low-Cost On-device Partial Domain Adaptation (LoCO-PDA): Enabling efficient CNN retraining on edge devicescited by 0
  6. 2023Efficient IoT Inference via Context-Awarenesscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2024Hollowed Net for On-Device Personalization of Text-to-Image Diffusion Models
  2. 2024Exploring the Boundaries of On-Device Inference: When Tiny Falls Short, Go Hierarchical
  3. 2024Block Selective Reprogramming for On-device Training of Vision Transformers
  4. 2025iOS as Acceleration
  5. 2024SmartMem: Layout Transformation Elimination and Adaptation for Efficient DNN Execution on Mobile
  6. 2023Speed Is All You Need: On-Device Acceleration of Large Diffusion Models via GPU-Aware Optimizations
  7. 2024Low-Energy On-Device Personalization for MCUs
  8. 2026MobileMoE: Scaling On-Device Mixture of Experts