AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Meta-GF: Training Dynamic-Depth Neural Networks Harmoniously

Yi Sun, Jian Li, Xin Xu

ECCV 2022 · Architectures and efficiency › RNNs and normalization · approach: Efficient network compression and training · 12 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
train with Meta-GFMSDnet · CIFAR-100average top-1 accuracy over exits: 72.83 -> 74.57Improved
train with Meta-GFMSDnet · CIFAR-10average top-1 accuracy over exits: 93.80 -> 94.54Improved
train with Meta-GFMSDnet · ImageNetaverage top-1 accuracy over exits: 66.72 -> 67.25Improved
train with Meta-GFSDN-VGG · CIFAR-100average top-1 accuracy over exits: 66.34 -> 68.51Improved
train with Meta-GFSDN-ResNet · CIFAR-100average top-1 accuracy over exits: 59.29 -> 60.32Improved
train with Meta-GF instead of CagradSDN-ResNet · CIFAR-10average top-1 accuracy over exits: 87.04 -> 86.62Hurt
train with Meta-GF vs CagradSDN-VGG · CIFAR-10average top-1 accuracy over exits: 89.54 -> 89.82Improved
train with Meta-GF instead of PcgradMSDnet · CIFAR-100average top-1 accuracy over exits: 74.14 -> 74.57Improved
add Meta-GF on shallow exitMSDnet · ImageNet exit 1average top-1 accuracy over exits: 58.48 -> 57.43Hurt
add Meta-GF on deepest exitMSDnet · ImageNet exit 5top-1 accuracy: 71.03 -> 73.27Improved
add knowledge distillation to Meta-GFMSDnet · CIFAR-100 (150 per class)average top-1 accuracy over exits: 57.93 -> 59.87Improved
train with Meta-GF vs baselineMSDnet · CIFAR-100 (150 per class)average top-1 accuracy over exits: 56.38 -> 57.93Improved

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