AI Paper Atlas

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

AAAI 2020 · Architectures and efficiency › Pruning and efficient tuning · approach: Low-rank compression and merging · 18 citations (OpenAlex)

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 changedWhereMeasuredEffect
Jointly optimize channel selection and model weights with trainable gatesResNet-56 · CIFAR-10accuracy: 91.9 -> 92.36Improved
Jointly optimize weight selection and model weights with trainable gatesVGG-16 · ImageNetaccuracy: 88.2 -> 88.18Improved
Prune channels to ρch = 0.1N-style transfer network · Style transferFile size (MB): 6.9 -> 0.2Improved
Prune channels to ρch = 0.1N-style transfer network · Style transferTime (ms): 549 -> 160Improved
Prune channels to ρch = 0.45FlowNetS · Flying Chairs validationEPE: 3.15 -> 3.13No clear effect
Prune attention heads to retain ρattn = 0.39Transformer · WMT 2014 English-to-German newstest2014BLEU: 27.32 -> 27.14Hurt
Use different derivative shaping functions g(w) instead of the constant g(w) = 1 (ablation)Trainable gate function · Image classification pruningaccuracyNo clear effect
Use deterministic trainable gates instead of probabilistic softmax selection20-node sine-function selection experiment · Synthetic sine-function taskretained hidden nodesImproved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy