AI Paper Atlas

Research map / Architectures and efficiency / CNN design

ARMA Nets: Expanding Receptive Field for Dense Prediction

Jiahao Su, Shiqi Wang, Furong Huang

NeurIPS 2020 · Architectures and efficiency › CNN design · approach: Efficient CNN scaling and dynamic inference · 0 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
Replace ConvLSTM convolutions with ARMA layersConvLSTM, kernel size 3 · Moving-MNIST-2PSNR: 18.24 -> 19.72Improved
Add non-local attention blocks (ablation)ConvLSTM, kernel size 5 · Moving-MNIST-2PSNR: 19.58 -> 19.18Hurt
Add non-local attention blocks (ablation)ConvLSTM, kernel size 3 · Moving-MNIST-2PSNR: 18.24 -> 19.45Improved
Add non-local attention blocks (ablation)ARMA-LSTM, kernel size 3 · Moving-MNIST-2PSNR: 19.72 -> 19.62No clear effect
Replace convolution layers with ARMA layersAlexNet · CIFAR10accuracy: 86.30 ± 0.29 -> 85.67 ± 0.19Hurt
Replace convolution layers with ARMA layersVGG-11 · CIFAR10accuracy: 91.57 ± 0.59 -> 91.57 ± 0.73No clear effect
Replace convolution layers with ARMA layersResNet-18 · CIFAR10accuracy: 95.01 ± 0.15 -> 95.07 ± 0.13No clear effect
Replace convolution layers with ARMA layersAlexNet · CIFAR100accuracy: 58.99 ± 0.37 -> 57.43 ± 0.24Hurt
Replace convolution layers with ARMA layersVGG-11 · CIFAR100accuracy: 68.25± 0.11 -> 68.36 ± 1.67No clear effect
Replace convolution layers with ARMA layersResNet-18 · CIFAR100accuracy: 73.71 ± 0.23 -> 73.72 ± 0.52No clear effect
Use stability-guaranteeing re-parameterization (ablation)ARMA-VGG-11 · CIFAR-10training convergenceImproved
Replace U-Net convolutions with ARMA layersU-Net · ISIC 2018 lesion segmentationsegmentation metricsImproved

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