# BlockDrop: Dynamic Inference Paths in Residual Networks

arXiv: https://arxiv.org/abs/1711.08393 · HTML: https://mlsanity.com/paper/1711.08393/

## Key facts

- Cited by 189 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 22 Nov 2017.

## Built on by (most cited first, 30 of 189)

- [Ego-Exo4D: Understanding Skilled Human Activity from First- and Third-Person Perspectives](https://mlsanity.com/paper/2311.18259/) (2023)
- [Dynamic Neural Networks: A Survey](https://mlsanity.com/paper/2102.04906/) (2021)
- [AdaViT: Adaptive Tokens for Efficient Vision Transformer](https://mlsanity.com/paper/2112.07658/) (2021)
- [Pruning and Quantization for Deep Neural Network Acceleration: A Survey](https://mlsanity.com/paper/2101.09671/) (2021)
- [AdaViT: Adaptive Vision Transformers for Efficient Image Recognition](https://mlsanity.com/paper/2111.15668/) (2021)
- [Are Multimodal Transformers Robust to Missing Modality?](https://mlsanity.com/paper/2204.05454/) (2022)
- [Not All Images are Worth 16x16 Words: Dynamic Transformers for Efficient Image Recognition](https://mlsanity.com/paper/2105.15075/) (2021)
- [IA-RED2: Interpretability-Aware Redundancy Reduction for Vision Transformers](https://mlsanity.com/paper/2106.12620/) (2021)
- [Learning Task-Oriented Communication for Edge Inference: An Information Bottleneck Approach](https://mlsanity.com/paper/2102.04170/) (2021)
- [MOOD: Multi-level Out-of-distribution Detection](https://mlsanity.com/paper/2104.14726/) (2021)
- [Spatially-Adaptive Image Restoration using Distortion-Guided Networks](https://mlsanity.com/paper/2108.08617/) (2021)
- [Adaptive Inference through Early-Exit Networks: Design, Challenges and Directions](https://mlsanity.com/paper/2106.05022/) (2021)
- [FrameExit: Conditional Early Exiting for Efficient Video Recognition](https://mlsanity.com/paper/2104.13400/) (2021)
- [A Survey on Green Deep Learning](https://mlsanity.com/paper/2111.05193/) (2021)
- [AdaFuse: Adaptive Temporal Fusion Network for Efficient Action Recognition](https://mlsanity.com/paper/2102.05775/) (2021)
- [AIM: Adaptive Inference of Multi-Modal LLMs via Token Merging and Pruning](https://mlsanity.com/paper/2412.03248/) (2024)
- [Glance and Focus Networks for Dynamic Visual Recognition](https://mlsanity.com/paper/2201.03014/) (2022)
- [SEENN: Towards Temporal Spiking Early-Exit Neural Networks](https://mlsanity.com/paper/2304.01230/) (2023)
- [Edge-Cloud Polarization and Collaboration: A Comprehensive Survey for AI](https://mlsanity.com/paper/2111.06061/) (2021)
- [AdaMML: Adaptive Multi-Modal Learning for Efficient Video Recognition](https://mlsanity.com/paper/2105.05165/) (2021)
- [Dynamic Network Quantization for Efficient Video Inference](https://mlsanity.com/paper/2108.10394/) (2021)
- [Efficient Split-Mix Federated Learning for On-Demand and In-Situ Customization](https://mlsanity.com/paper/2203.09747/) (2022)
- [SpaceByte: Towards Deleting Tokenization from Large Language Modeling](https://mlsanity.com/paper/2404.14408/) (2024)
- [Enable Deep Learning on Mobile Devices: Methods, Systems, and Applications](https://mlsanity.com/paper/2204.11786/) (2022)
- [Dynamic Resolution Network](https://mlsanity.com/paper/2106.02898/) (2021)
- [Dynamic Transfer for Multi-Source Domain Adaptation](https://mlsanity.com/paper/2103.10583/) (2021)
- [MicroNet: Improving Image Recognition with Extremely Low FLOPs](https://mlsanity.com/paper/2108.05894/) (2021)
- [Deep Joint Source-Channel Coding for Wireless Image Transmission with Adaptive Rate Control](https://mlsanity.com/paper/2110.04456/) (2021)
- [Task2Sim : Towards Effective Pre-training and Transfer from Synthetic Data](https://mlsanity.com/paper/2112.00054/) (2021)
- [Latency-aware Unified Dynamic Networks for Efficient Image Recognition](https://mlsanity.com/paper/2308.15949/) (2023)

## Abstract (arXiv)

Very deep convolutional neural networks offer excellent recognition results, yet their computational expense limits their impact for many real-world applications. We introduce BlockDrop, an approach that learns to dynamically choose which layers of a deep network to execute during inference so as to best reduce total computation without degrading prediction accuracy. Exploiting the robustness of Residual Networks (ResNets) to layer dropping, our framework selects on-the-fly which residual blocks to evaluate for a given novel image. In particular, given a pretrained ResNet, we train a policy network in an associative reinforcement learning setting for the dual reward of utilizing a minimal number of blocks while preserving recognition accuracy. We conduct extensive experiments on CIFAR and ImageNet. The results provide strong quantitative and qualitative evidence that these learned policies not only accelerate inference but also encode meaningful visual information. Built upon a ResNet-101 model, our method achieves a speedup of 20% on average, going as high as 36% for some images, while maintaining the same 76.4% top-1 accuracy on ImageNet.