Research map / Architectures and efficiency / RNNs and normalization
Dynamic Deep Neural Networks: Optimizing Accuracy-Efficiency Trade-Offs by Selective Execution
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use a chain D²NN instead of the Dynamic Capacity Network | Cluttered MNIST classification · Cluttered MNIST | multiplication cost: 2.77\times10^7 -> 2.66\times10^7 | Helped |
| Use a chain D²NN instead of the Dynamic Capacity Network | Cluttered MNIST classification · Cluttered MNIST | accuracy: 0.9861 -> 0.9698 | Hurt |
| Use a cascade D²NN to selectively execute computation | CIFAR-10 with ResNet-110 baseline · CIFAR-10 | cost reduction: 30\% reduction of cost with a 2\% loss (relative) on accuracy | Helped |
| Use a cascade D²NN to selectively execute computation more aggressively | CIFAR-10 with ResNet-110 baseline · CIFAR-10 | cost reduction: 62\% reduction of cost with a 7\% loss (relative) on accuracy | Helped |
| Optimize the cascade D²NN reward directly using wall time | Cascade D²NN · LFW-B | wall-time versus multiplication cost: 82\% wall-time cost corresponds to a 53\% number-of-multiplication cost; 95\% corresponds to 70\% | Helped |
Closest papers by meaning
- Adaptive Neural Networks for Efficient Inference (ICML 2017 · similarity 0.87)
- Dynamic Capacity Networks (ICML 2016 · similarity 0.87)
- ED-Batch: Efficient Automatic Batching of Dynamic Neural Networks via Learned Finite State Machines (ICML 2023 · similarity 0.86)
- Glance and Focus: a Dynamic Approach to Reducing Spatial Redundancy in Image Classification (NeurIPS 2020 · similarity 0.86)
- Recurrent neural network dynamical systems for biological vision (NeurIPS 2024 · similarity 0.85)
- Feature Learning Dynamics in Infinite-Depth Neural Networks (NeurIPS 2026 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.