Break the Ceiling: Stronger Multi-scale Deep Graph Convolutional Networks

5 Jun 2019Sitao LuanMingde ZhaoXiao-Wen ChangDoina Precup

Recently, neural network based approaches have achieved significant improvement for solving large, complex, graph-structured problems. However, their bottlenecks still need to be addressed, and the advantages of multi-scale information and deep architectures have not been sufficiently exploited... (read more)

PDF Abstract
Task Dataset Model Metric name Metric value Global rank Compare
Node Classification CiteSeer (0.5%) Snowball (linear) Accuracy 59.41% # 4
Node Classification CiteSeer (0.5%) Truncated Krylov Accuracy 64.64% # 1
Node Classification CiteSeer (0.5%) Snowball (tanh) Accuracy 62.05% # 2
Node Classification CiteSeer (0.5%) Snowball (linear + tanh) Accuracy 61.99% # 3
Node Classification CiteSeer (1%) Snowball (linear) Accuracy 65.85% # 3
Node Classification CiteSeer (1%) Snowball (tanh) Accuracy 64.23% # 4
Node Classification CiteSeer (1%) Snowball (linear + tanh) Accuracy 67.07% # 2
Node Classification CiteSeer (1%) Truncated Krylov Accuracy 69.03% # 1
Node Classification CiteSeer with Public Split: fixed 20 nodes per class Snowball (linear) Accuracy 73.11% # 3
Node Classification CiteSeer with Public Split: fixed 20 nodes per class Truncated Krylov Accuracy 73.58% # 2
Node Classification CiteSeer with Public Split: fixed 20 nodes per class Snowball (linear + tanh) Accuracy 72.55% # 6
Node Classification CiteSeer with Public Split: fixed 20 nodes per class Snowball (tanh) Accuracy 72.61% # 5
Node Classification Cora (0.5%) Snowball (linear) Accuracy 69.99% # 3
Node Classification Cora (0.5%) Snowball (linear + tanh) Accuracy 67.76% # 4
Node Classification Cora (0.5%) Snowball (tanh) Accuracy 71.36% # 2
Node Classification Cora (0.5%) Truncated Krylov Accuracy 74.89% # 1
Node Classification Cora (1%) Snowball (linear + tanh) Accuracy 74.79% # 2
Node Classification Cora (1%) Snowball (linear) Accuracy 73.10% # 4
Node Classification Cora (1%) Snowball (tanh) Accuracy 74.78% # 3
Node Classification Cora (1%) Truncated Krylov Accuracy 78.15% # 1
Node Classification Cora (3%) Snowball (linear) Accuracy 80.96% # 2
Node Classification Cora (3%) Snowball (linear + tanh) Accuracy 79.52% # 4
Node Classification Cora (3%) Snowball (tanh) Accuracy 80.72% # 3
Node Classification Cora (3%) Truncated Krylov Accuracy 81.92% # 1
Node Classification Cora with Public Split: fixed 20 nodes per class Snowball (linear) Accuracy 83.2% # 4
Node Classification Cora with Public Split: fixed 20 nodes per class Snowball (tanh) Accuracy 83.1% # 5
Node Classification Cora with Public Split: fixed 20 nodes per class Snowball (linear + tanh) Accuracy 82.8% # 7
Node Classification Cora with Public Split: fixed 20 nodes per class Truncated Krylov Accuracy 83.75% # 3
Node Classification PubMed (0.03%) Snowball (linear) Accuracy 68.12% # 2
Node Classification PubMed (0.03%) Snowball (linear + tanh) Accuracy 61.94% # 4
Node Classification PubMed (0.03%) Snowball (tanh) Accuracy 62.61% # 3
Node Classification PubMed (0.03%) Truncated Krylov Accuracy 71.11% # 1
Node Classification PubMed (0.05%) Truncated Krylov Accuracy 72.57% # 1
Node Classification PubMed (0.05%) Snowball (linear) Accuracy 70.04% # 2
Node Classification PubMed (0.05%) Snowball (linear + tanh) Accuracy 69.45% # 3
Node Classification PubMed (0.05%) Snowball (tanh) Accuracy 68.99% # 4
Node Classification PubMed (0.1%) Snowball (linear) Accuracy 73.83% # 4
Node Classification PubMed (0.1%) Snowball (linear + tanh) Accuracy 75.30% # 2
Node Classification PubMed (0.1%) Snowball (tanh) Accuracy 74.40% # 3
Node Classification PubMed (0.1%) Truncated Krylov Accuracy 77.21% # 1
Node Classification PubMed with Public Split: fixed 20 nodes per class Snowball (tanh) Accuracy 78.95% # 6
Node Classification PubMed with Public Split: fixed 20 nodes per class Snowball (linear + tanh) Accuracy 79.04% # 4
Node Classification PubMed with Public Split: fixed 20 nodes per class Truncated Krylov Accuracy 80.01% # 1
Node Classification PubMed with Public Split: fixed 20 nodes per class Snowball (linear) Accuracy 79.23% # 3