AI Paper Atlas

Research map / Graph learning / GNN architectures

Learning Connectivity of Neural Networks from a Topological Perspective

Kun Yuan, Quanquan Li, Jing Shao, Junjie Yan

ECCV 2020 · Graph learning › GNN architectures · approach: Theory of graph convolutional networks · 2 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
add learnable edge weights alpha to the complete graph (ablation)TopoNet complete · ImageNet valTop-1 accuracy: 77.24 -> 78.22Improved
add uniform sparsity constraint on alpha (ablation)TopoNet complete + alpha · ImageNet valTop-1 accuracy: 78.22 -> 78.46Improved
add adaptive (in-degree log-scaled) sparsity constraint on alpha (ablation)TopoNet complete + alpha · ImageNet valTop-1 accuracy: 78.22 -> 78.60Improved
replace uniform with adaptive sparsity (ablation)TopoNet complete + alpha · ImageNet valTop-1 accuracy: 78.46 -> 78.60Improved
learn connectivity on ResNet-110 (complete graph, interval 1)ResNet-110 · CIFAR-100 valTop-1 accuracy: 76.31 -> 78.54Improved
learn connectivity on ResNet-20 (smaller search space)ResNet-20 · CIFAR-100 valTop-1 accuracy: 69.01 -> 69.91Improved
learn connectivity on MobileNetV2-1.0 baseline depthMobileNetV2-1.0 · ImageNet valTop-1 accuracy: 72.60 -> 72.86Improved
learn connectivity on 6x deeper MobileNetV2 (larger search space)MobileNetV2-1.0-6N · ImageNet valTop-1 accuracy: 77.61 -> 78.36Improved
replace residual l=2 topology with learned complete topology with adaptive sparsityTopoNet · COCO val2017AP: 40.74 -> 41.69Improved
replace ResNet-50 backbone with learned TopoNet backboneFPN · COCO val2017AP: 36.42 -> 41.69Improved
reduce residual interval from 4 to 2 (denser connections) (ablation)TopoNet residual · ImageNet valTop-1 accuracy: 77.72 -> 78.26Improved
make the complete graph random with p=0.4 instead of fully connected (ablation)TopoNet · ImageNet valTop-1 accuracy: 77.24 -> 77.90Improved

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