Code implementation of the paper: Graph Stochastic Neural Networks for Semi-supervised Learning, which has been accepted by NeurIPS 2020.

• python 3.6.7
• numpy 1.15.4
• scipy 1.1.0
• scikit-learn 0.20.2
• matplotlib 3.0.2
• torch 1.1.0
• tqdm 4.31.1

All experiments are conducted on a server with the following configurations:

• Operating System: CentOS Linux release 7.4.1708
• CPU: Intel(R) Xeon(R) CPU E5-2620 v4 @ 2.20GHz
• GPU: GeForce GTX TITAN X

To try our code, you can use the IPython notebook demo1_standard.ipynb, demo2_scarcelabel.ipynb and demo3_attack.ipynb for three different experimental scenarios.

• Standard Experimental Scenario: demo1_standard.ipynb
• Label-Scarce Scenario: demo2_scarcelabel.ipynb
• Adversarial Attack Scenario: demo3_attack.ipynb

In the folder ./data, we provide the following datasets:

Besides, to evaluate the performance of GSNN in the presence of adversarial attacks, we also provide the following poisoned graphs for Cora generated by different attack methods in the folder ./data (attack budget: 0.05).

cora_0.05edges_Meta-Self.npy

cora_0.05edges_Meta-Train.npy

cora_0.05_minmax.npy