mlsanityData to 7 Oct 2026

Adaptive Neural Networks for Efficient Inference

arXiv 1702.07811 ↗First posted 25 Feb 2017cs.LG cs.CV cs.NEcited by 161

Not on the map. Why?

Abstract (arXiv)

We present an approach to adaptively utilize deep neural networks in order to reduce the evaluation time on new examples without loss of accuracy. Rather than attempting to redesign or approximate existing networks, we propose two schemes that adaptively utilize networks. We first pose an adaptive network evaluation scheme, where we learn a system to adaptively choose the components of a deep network to be evaluated for each example. By allowing examples correctly classified using early layers of the system to exit, we avoid the computational time associated with full evaluation of the network. We extend this to learn a network selection system that adaptively selects the network to be evaluated for each example. We show that computational time can be dramatically reduced by exploiting the fact that many examples can be correctly classified using relatively efficient networks and that complex, computationally costly networks are only necessary for a small fraction of examples. We pose a global objective for learning an adaptive early exit or network selection policy and solve it by reducing the policy learning problem to a layer-by-layer weighted binary classification problem. Empirically, these approaches yield dramatic reductions in computational cost, with up to a 2.8x speedup on state-of-the-art networks from the ImageNet image recognition challenge with minimal (<1%) loss of top5 accuracy.

Key facts. Cited by 161 papers in the map (data through 7 Oct 2026). First posted on arXiv 25 Feb 2017.

Builds on

No reference list yet.

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2019: 1; 2020: 0; 2021: 22; 2022: 26; 2023: 38; 2024: 26; 2025: 26; 2026: 22.

161 papers.

Sort:

By area:

  1. 2022Measuring and Narrowing the Compositionality Gap in Language Modelscited by 1,054
  2. 2021Dynamic Neural Networks: A Surveycited by 405
  3. 2021Pruning and Quantization for Deep Neural Network Acceleration: A Surveycited by 259
  4. 2021PonderNet: Learning to Pondercited by 156
  5. 2024DeeR-VLA: Dynamic Inference of Multimodal Large Language Models for Efficient Robot Executioncited by 116
  6. 2024Not All Layers of LLMs Are Necessary During Inferencecited by 95
  7. 2021MOOD: Multi-level Out-of-distribution Detectioncited by 94
  8. 2024Dynamic Diffusion Transformercited by 60
  9. 2023Towards Accurate Image Coding: Improved Autoregressive Image Generation with Dynamic Vector Quantizationcited by 59
  10. 2022Dynamic Multimodal Fusioncited by 54
  11. 2023A Dynamic Multi-Scale Voxel Flow Network for Video Predictioncited by 53
  12. 2021A Survey on Green Deep Learningcited by 52
38 more
  1. 2023Adaptive Rotated Convolution for Rotated Object Detectioncited by 51
  2. 2022Learning to Weight Samples for Dynamic Early-exiting Networkscited by 44
  3. 2021Shapley Explanation Networkscited by 41
  4. 2022Glance and Focus Networks for Dynamic Visual Recognitioncited by 41
  5. 2021Edge-Cloud Polarization and Collaboration: A Comprehensive Survey for AIcited by 38
  6. 2021Teachers Do More Than Teach: Compressing Image-to-Image Modelscited by 37
  7. 2024Learning Iterative Reasoning through Energy Diffusioncited by 36
  8. 2023Dynamic Perceiver for Efficient Visual Recognitioncited by 33
  9. 2022Learning Iterative Reasoning through Energy Minimizationcited by 32
  10. 2023Learning to Reason and Memorize with Self-Notescited by 30
  11. 2023CoTFormer: A Chain-of-Thought Driven Architecture with Budget-Adaptive Computation Cost at Inferencecited by 27
  12. 2024Fast yet Safe: Early-Exiting with Risk Controlcited by 27
  13. 2024Dynamic Tuning Towards Parameter and Inference Efficiency for ViT Adaptationcited by 25
  14. 2022Controllable Dynamic Multi-Task Architecturescited by 21
  15. 2025Continuous Thought Machinescited by 21
  16. 2023AdaptiveNet: Post-deployment Neural Architecture Adaptation for Diverse Edge Environmentscited by 19
  17. 2021Multi-Exit Semantic Segmentation Networkscited by 18
  18. 2022Scalable Video Object Segmentation with Identification Mechanismcited by 18
  19. 2022Efficient Deep Visual and Inertial Odometry with Adaptive Visual Modality Selectioncited by 17
  20. 2021AppealNet: An Efficient and Highly-Accurate Edge/Cloud Collaborative Architecture for DNN Inferencecited by 16
  21. 2022DeepPerform: An Efficient Approach for Performance Testing of Resource-Constrained Neural Networkscited by 16
  22. 2024Agreement-Based Cascading for Efficient Inferencecited by 16
  23. 2023Jointly-Learned Exit and Inference for a Dynamic Neural Network : JEI-DNNcited by 15
  24. 2023Apparate: Rethinking Early Exits to Tame Latency-Throughput Tensions in ML Servingcited by 15
  25. 2025How Do LLMs Use Their Depth?cited by 15
  26. 2023Computation-efficient Deep Learning for Computer Vision: A Surveycited by 14
  27. 2022DSE-GAN: Dynamic Semantic Evolution Generative Adversarial Network for Text-to-Image Generationcited by 13
  28. 2023Real-Time Image Demoireing on Mobile Devicescited by 13
  29. 2024LF-ViT: Reducing Spatial Redundancy in Vision Transformer for Efficient Image Recognitioncited by 13
  30. 2022Federated learning: Applications, challenges and future directionscited by 11
  31. 2023Input-Aware Dynamic Timestep Spiking Neural Networks for Efficient In-Memory Computingcited by 11
  32. 2024Conditional computation in neural networks: principles and research trendscited by 10
  33. 2025Gatekeeper: Improving Model Cascades Through Confidence Tuningcited by 10
  34. 2021Dynamic Slimmable Denoising Networkcited by 9
  35. 2026AdaPonderLM: Gated Pondering Language Models with Token-Wise Adaptive Depthcited by 9
  36. 2021E2CM: Early Exit via Class Means for Efficient Supervised and Unsupervised Learningcited by 8
  37. 2021When in Doubt, Summon the Titans: Efficient Inference with Large Modelscited by 8
  38. 2023AsConvSR: Fast and Lightweight Super-Resolution Network with Assembled Convolutionscited by 8

50 most cited of 161 All 161 (JSON)