AI Paper Atlas

Research map / Graph learning / GNN architectures

Digraph Inception Convolutional Networks

Zekun Tong, Yuxuan Liang, Changsheng Sun, Xinke Li, David Rosenblum, Andrew Lim

NeurIPS 2020 · Graph learning › GNN architectures · approach: Deep graph convolution and simplification · 34 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 Inception block (k-th order proximity) to digraph convolution (ablation)DiGCN w/ Lappr · CORA-MLaccuracy (%): 77.01 -> 80.28Improved
add Inception block (k-th order proximity) to digraph convolution (ablation)DiGCN w/ Lappr · CITESEERaccuracy (%): 64.92 -> 66.11Improved
add Inception block (k-th order proximity) to digraph convolution (ablation)DiGCN w/ Lappr · AM-PHOTOaccuracy (%): 88.72 -> 90.02Improved
add Inception block (k-th order proximity) to digraph convolution (ablation)DiGCN w/ Lappr · AM-COMPUTERaccuracy (%): 85.55 -> 85.94Improved
use sparse personalized-PageRank Laplacian Lappr instead of dense PageRank Laplacian Lpr (ablation)DiGCN w/o IB · CORA-MLaccuracy (%): 77.11 -> 77.01No clear effect
use sparse personalized-PageRank Laplacian Lappr instead of dense Lpr (ablation)DiGCN w/o IB · AM-PHOTOaccuracy (%): OOM -> 88.72Improved
use directed PageRank digraph convolution instead of undirected GCNGCN · CORA-MLaccuracy (%): 53.11 -> 77.01Improved
increase Inception layers from 1 to 3 (ablation)DiGCN · CORA-MLaccuracy (%): 72.14 -> 80.28Improved
increase Inception layers from 3 to 4 (ablation)DiGCN · CORA-MLaccuracy (%): 80.28 -> 79.95No clear effect
increase model width k beyond 2 (ablation)DiGCN single layer · CITESEERaccuracy (%)No clear effect
sum vs concatenation fusion (ablation)DiGCN · AM-PHOTOaccuracy (%)Mixed
use ReLU non-linear activation in Inception block (ablation)DiGCN · AM-PHOTOaccuracy (%)Hurt

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