[InfoGraph] Getting different results from the reported

[galinslp]

Hi, I performed some experiments with the supervised baseline of the Infograph (https://github.com/fanyun-sun/InfoGraph/tree/master/semi-supervised), and I got a result different from the reported in the last table in your README.md file (https://github.com/dmlc/dgl/tree/master/examples/pytorch/infograph). For example, for the mu (0) property, I got a test error, in terms of MAE, of 0.2285302, while in your table, the reported error is 0.3169.

So, I have two questions:

• The supervised results errors were obtained using the supervised baseline available in https://github.com/fanyun-sun/InfoGraph/tree/master/semi-supervised?

• How can I obtain the same results reported in the table in https://github.com/dmlc/dgl/tree/master/examples/pytorch/infograph?

Thanks!!

[hengruz]

Hi galinslp,
Actually, for mu(0) property, 0.3169 is that reported in the paper, 0.2411 is what we try using the author’s code: InfoGraph/semi-supervised at master · fanyun-sun/InfoGraph · GitHub, and 0.2355 is our implementation using DGL: dgl/examples/pytorch/infograph at master · dmlc/dgl · GitHub.