# Deciding How to Decide: Dynamic Routing in Artificial Neural Networks

arXiv: https://arxiv.org/abs/1703.06217 · HTML: https://mlsanity.com/paper/1703.06217/

## Key facts

- Cited by 33 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 17 Mar 2017.

## Built on by (most cited first, 30 of 33)

- [Dynamic Neural Networks: A Survey](https://mlsanity.com/paper/2102.04906/) (2021)
- [AdaFuse: Adaptive Temporal Fusion Network for Efficient Action Recognition](https://mlsanity.com/paper/2102.05775/) (2021)
- [Zero Time Waste: Recycling Predictions in Early Exit Neural Networks](https://mlsanity.com/paper/2106.05409/) (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)
- [VORTA: Efficient Video Diffusion via Routing Sparse Attention](https://mlsanity.com/paper/2505.18809/) (2025)
- [Anytime Dense Prediction with Confidence Adaptivity](https://mlsanity.com/paper/2104.00749/) (2021)
- [TESTAM: A Time-Enhanced Spatio-Temporal Attention Model with Mixture of Experts](https://mlsanity.com/paper/2403.02600/) (2024)
- [Neural Networks are Decision Trees](https://mlsanity.com/paper/2210.05189/) (2022)
- [VA-RED2: Video Adaptive Redundancy Reduction](https://mlsanity.com/paper/2102.07887/) (2021)
- [DynamicDet: A Unified Dynamic Architecture for Object Detection](https://mlsanity.com/paper/2304.05552/) (2023)
- [Spatial Mixture-of-Experts](https://mlsanity.com/paper/2211.13491/) (2022)
- [Revisiting Single-gated Mixtures of Experts](https://mlsanity.com/paper/2304.05497/) (2023)
- [E2CM: Early Exit via Class Means for Efficient Supervised and Unsupervised Learning](https://mlsanity.com/paper/2103.01148/) (2021)
- [Boosted Dynamic Neural Networks](https://mlsanity.com/paper/2211.16726/) (2022)
- [AsConvSR: Fast and Lightweight Super-Resolution Network with Assembled Convolutions](https://mlsanity.com/paper/2305.03387/) (2023)
- [A Survey of Neural Trees](https://mlsanity.com/paper/2209.03415/) (2022)
- [ODAR: Principled Adaptive Routing for LLM Reasoning via Active Inference](https://mlsanity.com/paper/2602.23681/) (2026)
- [Understanding the Robustness of Multi-Exit Models under Common Corruptions](https://mlsanity.com/paper/2212.01562/) (2022)
- [EgoAdapt: Adaptive Multisensory Distillation and Policy Learning for Efficient Egocentric Perception](https://mlsanity.com/paper/2506.21080/) (2025)
- [Sparsely ensembled convolutional neural network classifiers via reinforcement learning](https://mlsanity.com/paper/2102.03921/) (2021)
- [Complexity-aware Adaptive Training and Inference for Edge-Cloud Distributed AI Systems](https://mlsanity.com/paper/2109.06440/) (2021)
- [ATESA-BAERT: A Heterogeneous Ensemble Learning Model for Aspect-Based Sentiment Analysis](https://mlsanity.com/paper/2307.15920/) (2023)
- [Divide, Specialize, and Route: A New Approach to Efficient Ensemble Learning](https://mlsanity.com/paper/2506.20814/) (2025)
- [Transition Matrix Representation of Trees with Transposed Convolutions](https://mlsanity.com/paper/2202.10677/) (2022)
- [Neural Routing in Meta Learning](https://mlsanity.com/paper/2210.07932/) (2022)
- [Unifying Synergies between Self-supervised Learning and Dynamic Computation](https://mlsanity.com/paper/2301.09164/) (2023)
- [LEURN: Learning Explainable Univariate Rules with Neural Networks](https://mlsanity.com/paper/2303.14937/) (2023)
- [Conditional Information Gain Trellis](https://mlsanity.com/paper/2402.08345/) (2024)
- [Rule Based Learning with Dynamic (Graph) Neural Networks](https://mlsanity.com/paper/2406.09954/) (2024)

## Abstract (arXiv)

We propose and systematically evaluate three strategies for training dynamically-routed artificial neural networks: graphs of learned transformations through which different input signals may take different paths. Though some approaches have advantages over others, the resulting networks are often qualitatively similar. We find that, in dynamically-routed networks trained to classify images, layers and branches become specialized to process distinct categories of images. Additionally, given a fixed computational budget, dynamically-routed networks tend to perform better than comparable statically-routed networks.