Research map / Architectures and efficiency / Pruning and efficient tuning
Plug-in, Trainable Gate for Streamlining Arbitrary Neural Networks
Jaedeok Kim, Chiyoun Park, Hyun-Joo Jung, Yoonsuck Choe
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| Jointly optimize channel selection and model weights with trainable gates | ResNet-56 · CIFAR-10 | accuracy: 91.9 -> 92.36 | Improved |
| Jointly optimize weight selection and model weights with trainable gates | VGG-16 · ImageNet | accuracy: 88.2 -> 88.18 | Improved |
| Prune channels to ρch = 0.1 | N-style transfer network · Style transfer | File size (MB): 6.9 -> 0.2 | Improved |
| Prune channels to ρch = 0.1 | N-style transfer network · Style transfer | Time (ms): 549 -> 160 | Improved |
| Prune channels to ρch = 0.45 | FlowNetS · Flying Chairs validation | EPE: 3.15 -> 3.13 | No clear effect |
| Prune attention heads to retain ρattn = 0.39 | Transformer · WMT 2014 English-to-German newstest2014 | BLEU: 27.32 -> 27.14 | Hurt |
| Use different derivative shaping functions g(w) instead of the constant g(w) = 1 (ablation) | Trainable gate function · Image classification pruning | accuracy | No clear effect |
| Use deterministic trainable gates instead of probabilistic softmax selection | 20-node sine-function selection experiment · Synthetic sine-function task | retained hidden nodes | Improved |
Closest papers by meaning
- GDP: Stabilized Neural Network Pruning via Gates With Differentiable Polarization (ICCV 2021 · similarity 0.86)
- Discrete Model Compression With Resource Constraint for Deep Neural Networks (CVPR 2020 · similarity 0.84)
- Gate Decorator: Global Filter Pruning Method for Accelerating Deep Convolutional Neural Networks (NeurIPS 2019 · similarity 0.84)
- A Signal Propagation Perspective for Pruning Neural Networks at Initialization (ICLR 2020 · similarity 0.84)
- Rethinking Neural Nonlinearity as Gating (NeurIPS 2026 · similarity 0.82)
- Auto-Train-Once: Controller Network Guided Automatic Network Pruning from Scratch (CVPR 2024 · similarity 0.82)
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.
