BOND data possible inconsistency

[realfolkcode]

Describe the bug In the BOND paper, it is said that all the datasets are undirected, except Weibo.

Note that Weibo is a directed graph; the remaining datasets used in our benchmark are undirected graphs.

However, load_data function returns directed PyG graphs (only "reddit" is undirected for some reason). Here is the output of is_undirected method

inj_cora False
inj_amazon False
inj_flickr False
weibo False
reddit True
disney False
books False
enron False```

To Reproduce Here is a colab notebook to reproduce the output above https://colab.research.google.com/drive/1mNXh66Ac2hUduHvzKtGifC7_huBgCf-5?usp=sharing

Expected behavior I expected the data to be consistent with what is stated in the paper. Please let me know if I misunderstood something or it's indeed a mistake. Thanks!

[yzfxmu]

Describe the bug In the BOND paper, it is said that all the datasets are undirected, except Weibo.

Note that Weibo is a directed graph; the remaining datasets used in our benchmark are undirected graphs.

However, load_data function returns directed PyG graphs (only "reddit" is undirected for some reason). Here is the output of is_undirected method

inj_cora False
inj_amazon False
inj_flickr False
weibo False
reddit True
disney False
books False
enron False```

To Reproduce Here is a colab notebook to reproduce the output above https://colab.research.google.com/drive/1mNXh66Ac2hUduHvzKtGifC7_huBgCf-5?usp=sharing

Expected behavior I expected the data to be consistent with what is stated in the paper. Please let me know if I misunderstood something or it's indeed a mistake. Thanks!

PYG treat all graph as directed. In this case, the datasets mentioned above have edges from two directions.

[realfolkcode]

PYG treat all graph as directed. In this case, the datasets mentioned above have edges from two directions.

@yzfxmu it is true that PyG's implementation of graphs is directed. As you correctly pointed out, for undirected graphs, their edges are considered bidirected. Therefore, each edge should appear twice in edge_index (assuming there are no self-loops). However, this is not the case for most of the graphs in BOND, as can be checked by is_undirected method. The possible consequence of this is suboptimal performance of message passing layers.

[kayzliu]

Thanks for pointing out the problem. After carefully reviewing the datasets, we summarize the results as follows:

• Cora, Amazon, Flickr: the original graphs without injected outliers is undirected. During the injection of structural outliers, we should follow to add directed/undirected edges if the graph is directed/undirected. However, we mistakenly added directed edges. We have added a directed parameter in pygod.generator.gen_structural_outlier to enable adding directed or undirected edges in the future.
• Weibo, Reddit: these two datasets are consistent with the description in the paper.
• Disney, Books: these two datasets are undirected in their semantic meaning. torch_geometric.transforms.ToUndirected in can help us convert it to an undirected graph.
• Enron: this dataset should be directed.

In summary, apologize for the confusion made by our mistakes, but we are not able to change the datasets as we have reported the results based on these datasets in the paper. We provide potential solutions for each dataset.