search

sunfanyunn / InfoGraph

Official code for "InfoGraph: Unsupervised and Semi-supervised Graph-Level Representation Learning via Mutual Information Maximization" (ICLR 2020, spotlight)
https://openreview.net/forum?id=r1lfF2NYvH
312 stars 45 forks source link

Question about the reported results #4

Closed ha-lins closed 4 years ago

ha-lins commented 4 years ago

Hi @fanyun-sun,

Thanks for your released code! I have two minor questions about the reported results as follows:

  1. When I ran the code, the saved log file reported the performance of 20 epochs. I wonder how to obtain the reported results in the paper (e.g. 89.10 for MUTAG dataset).

Here are the printed results for MUTAG dataset: "linearsvc": [0.7950584795321638, 0.8650584795321636, 0.8351169590643275, 0.868216374269006, 0.8592397660818714, 0.8723391812865497, 0.8939766081871345, 0.8783625730994153, 0.8770175438596493, 0.855877192982456, 0.8981286549707603, 0.8586549707602339, 0.8731578947368421, 0.8789473684210527, 0.8550584795321639, 0.8631286549707602, 0.8656432748538011, 0.8717251461988305, 0.8503508771929825, 0.8814327485380117, 0.8831286549707602],

  1. Could you pls explain the meaning of 10-fold cross-validation accuracy for me? Maybe you can point out the specific implementation in the code.

Thanks a lot!