gemslab/h2gcn

Boost learning for GNNs from the graph structure under challenging heterophily settings. (NeurIPS'20)

h2gcn's Issues

about downloading dataset

thanks for your impressive work!
When I download the dataset, it seems that I can't connect to https://umich.app.box.com. Would you like to give me some suggestion about it ?
thx!
`--2021-01-11 19:04:34-- (try: 4) https://umich.app.box.com/public/static/oerjpreqd1u1cn481mk1m788fd4ahcvv.gz

Connecting to umich.app.box.com (umich.app.box.com)|2001::42dc:932f|:443... failed: Connection timed out.
Retrying.`

How to run on graphs with heterophily?

Hi,
I'd like to ask how to run the code on graphs with heterophily? e.g., Texas, Actor

How to get and use the node features of the syn-cora dataset?

Recently I had the pleasure of reading your paper "Beyond Homophily in Graph Neural Networks: Current Limitations and Effective Designs" published in NIPS and was impressed with the algorithm and experimental results. I am following the instructions on your github to try to reproduce your experiment using the H2GCN algorithm alone (instead of signac). I downloaded the synthetic dataset using the "get-syn-cora.sh" script file you provided, but during the experiment I noticed that the downloaded ".allx" file does not seem to contain the node features. (The data in ".allx" are some sparse, binary-valued matrices, although the format is ture.) Therefore, I would like to check with you: Does the downloaded dataset named "workspace" already include node features? If yes, how do I use it? Is it possible to call it directly using the same interface as Kipf? If not, how should I regenerate the features to be able to reproduce the results reported in Figure2(a)? Could you please provide it in the same as Kipf et.al, I believe it would be very profitable for the community and to promote your work. We sincerely look forward to hearing from you!

Can not run H2GCN on Actor, WebKB and Wikipedia Dataset

Hello,

According to your repository you have provided the citation (cora, citeseer, pubmed) datasets. I want to reproduce the results of H2GCN in Actor, WebKB (Cornell, Texas, Wisconsin) and Wikipedia (Chameleon, Squirrel) datasets.

But the citation datasets are transformed into .x, .y, .allx, .ally, .graph, .test.index in these format while other datasets mentioned above are not available in internet in these format.

Can you please provide the utility code which you used to load these datasets to run H2GCN on them?

ImportError: cannot import name 'FlowProject' from 'flow'

While trying to recreate experiments from experiments/h2gcn/
I am encountering this issue
ImportError: cannot import name 'FlowProject' from 'flow'
I cannot resolve it; can you help?

run H2GCN alone on graph with heterophily

Sorry to bother you with another question: How can I run H2GCN alone on graph with heterophily? After downloading them, how to load them without signac?

About heterophilic datasets

Firstly, thanks for your impressive work and the source code. I downloaded the datasets based on your script. For heterophilic datasets, I found there are two files in data_source folder (one for graph edges and one for node&feature&label). I would like to know are these two files the same as here (https://github.com/graphdml-uiuc-jlu/geom-gcn/tree/master/new_data)? Thanks!

About downloading geom-gcn datasets

Hi, I cannot download geom-gcn datasets (texas, Wisconsin, Cornell) using provided scripts.
Can you help?

How to run baseline model?

Hi,
I want to run baseline models with

cd  experiments/h2gcn  
python run_graphsage_experiments.py -c configs/real-geomgcn/graphsage.json run --debug

I got the following error:

[07-03 04:53:12 run_model@4c95f334af46cefe80977d9c96123445     INFO] Traceback (most recent call last):
[07-03 04:53:12 run_model@4c95f334af46cefe80977d9c96123445     INFO] File "model.py", line 10, in <module>
[07-03 04:53:12 run_model@4c95f334af46cefe80977d9c96123445     INFO] import dataset
[07-03 04:53:12 run_model@4c95f334af46cefe80977d9c96123445     INFO] ModuleNotFoundError: No module named 'dataset'
[07-03 04:53:13 run_model@4c95f334af46cefe80977d9c96123445    ERROR] Check log at /H2GCN/experiments/h2gcn/workspace/4c95f334af46cefe80977d9c96123445/features/e9cf54653de9231697763f2d1216eb5c/splits/eb8f45d8594346aa36c4773fc43f13fb/experiments/graphsage_experiments/terminal_output.log

And the terminal_output.log contains:

[07-03 04:43:07     INFO] ===============
>>>>Executing command ['/home/anaconda3/bin/python', '-u', 'model.py', '--dataset_path', '/H2GCN/experiments/h2gcn/workspace/4c95f334af46cefe80977d9c96123445/features/e9cf54653de9231697763f2d1216eb5c/splits/eb8f45d8594346aa36c4773fc43f13fb', '--dataset', 'ind.cora-unmodified-0.48p__0.2p', '--run_id=H2GCN --network_setup M64-T1-G-V-C1-MO --adj_nhood 1 2 --l2_regularize_weight 1e-5@ce7091201460a2a7d3384b919b781b63_9caf91da0e68a2cb7a0d7b4c648929c4_20c7d1b09c28a20e6fa374cc6c8544e8_8b2d686c68a5e9f7ff474419f73f3a6d_0c634e96f738e8800231132e2848e303_248ec654a2e8d5f30df7ac8ef12d3cf5_a4007892967f684b0be93efd41b1d2f3_83150345b828a1192e923bc2e926b66b', '--use_signac', '--signac_root', '/H2GCN/experiments/h2gcn/workspace/4c95f334af46cefe80977d9c96123445/features/e9cf54653de9231697763f2d1216eb5c/splits/eb8f45d8594346aa36c4773fc43f13fb/experiments/graphsage_experiments', '--val_size', '1019', 'H2GCN', '--network_setup', 'M64-T1-G-V-C1-MO', '--adj_nhood', '1', '2', '--l2_regularize_weight', '1e-5']
===============
[07-03 04:43:08     INFO] Traceback (most recent call last):
[07-03 04:43:08     INFO] File "model.py", line 10, in <module>
[07-03 04:43:08     INFO] import dataset
[07-03 04:43:08     INFO] ModuleNotFoundError: No module named 'dataset'
[07-03 04:43:08    ERROR] Check log at /H2GCN/experiments/h2gcn/workspace/4c95f334af46cefe80977d9c96123445/features/e9cf54653de9231697763f2d1216eb5c/splits/eb8f45d8594346aa36c4773fc43f13fb/experiments/graphsage_experiments/terminal_output.log
[07-03 04:45:57     INFO] ===============
>>>>Executing command ['/home/anaconda3/bin/python', '-u', 'model.py', '--dataset_path', ’/H2GCN/experiments/h2gcn/workspace/4c95f334af46cefe80977d9c96123445/features/e9cf54653de9231697763f2d1216eb5c/splits/eb8f45d8594346aa36c4773fc43f13fb', '--dataset', 'ind.cora-unmodified-0.48p__0.2p', '--run_id=H2GCN --network_setup M64-T1-G-V-C1-MO --adj_nhood 1 2 --l2_regularize_weight 1e-5@ce7091201460a2a7d3384b919b781b63_9caf91da0e68a2cb7a0d7b4c648929c4_20c7d1b09c28a20e6fa374cc6c8544e8_8b2d686c68a5e9f7ff474419f73f3a6d_0c634e96f738e8800231132e2848e303_248ec654a2e8d5f30df7ac8ef12d3cf5_a4007892967f684b0be93efd41b1d2f3_83150345b828a1192e923bc2e926b66b', '--use_signac', '--signac_root', '/H2GCN/experiments/h2gcn/workspace/4c95f334af46cefe80977d9c96123445/features/e9cf54653de9231697763f2d1216eb5c/splits/eb8f45d8594346aa36c4773fc43f13fb/experiments/graphsage_experiments', '--val_size', '1019', 'H2GCN', '--network_setup', 'M64-T1-G-V-C1-MO', '--adj_nhood', '1', '2', '--l2_regularize_weight', '1e-5']
===============
[07-03 04:45:58     INFO] Traceback (most recent call last):
[07-03 04:45:58     INFO] File "model.py", line 10, in <module>
[07-03 04:45:58     INFO] import dataset
[07-03 04:45:58     INFO] ModuleNotFoundError: No module named 'dataset'
[07-03 04:45:58    ERROR] Check log at /H2GCN/experiments/h2gcn/workspace/4c95f334af46cefe80977d9c96123445/features/e9cf54653de9231697763f2d1216eb5c/splits/eb8f45d8594346aa36c4773fc43f13fb/experiments/graphsage_experiments/terminal_output.log
[07-03 04:53:11     INFO] ===============
>>>>Executing command ['/anaconda3/bin/python', '-u', 'model.py', '--dataset_path', '/H2GCN/experiments/h2gcn/workspace/4c95f334af46cefe80977d9c96123445/features/e9cf54653de9231697763f2d1216eb5c/splits/eb8f45d8594346aa36c4773fc43f13fb', '--dataset', 'ind.cora-unmodified-0.48p__0.2p', '--run_id=--hid_units 64 --epochs 500@ce7091201460a2a7d3384b919b781b63_9caf91da0e68a2cb7a0d7b4c648929c4_20c7d1b09c28a20e6fa374cc6c8544e8_8b2d686c68a5e9f7ff474419f73f3a6d_0c634e96f738e8800231132e2848e303_248ec654a2e8d5f30df7ac8ef12d3cf5_a4007892967f684b0be93efd41b1d2f3_83150345b828a1192e923bc2e926b66b', '--use_signac', '--signac_root', '/H2GCN/experiments/h2gcn/workspace/4c95f334af46cefe80977d9c96123445/features/e9cf54653de9231697763f2d1216eb5c/splits/eb8f45d8594346aa36c4773fc43f13fb/experiments/graphsage_experiments', '--val_size', '1019', '--hid_units', '64', '--epochs', '500']
===============
[07-03 04:53:12     INFO] Traceback (most recent call last):
[07-03 04:53:12     INFO] File "model.py", line 10, in <module>
[07-03 04:53:12     INFO] import dataset
[07-03 04:53:12     INFO] ModuleNotFoundError: No module named 'dataset'
[07-03 04:53:13    ERROR] Check log at /H2GCN/experiments/h2gcn/workspace/4c95f334af46cefe80977d9c96123445/features/e9cf54653de9231697763f2d1216eb5c/splits/eb8f45d8594346aa36c4773fc43f13fb/experiments/graphsage_experiments/terminal_output.log

I don't know whether it is the right way to directly run the baseline models, could you help me solve this problem? Thx

About downloading syn-cora

Hi,
I try to access to https://umich.app.box.com/public/static/oerjpreqd1u1cn481mk1m788fd4ahcvv.gz. But it says 'The shared file or folder has been removed.' How to achieve the syn-cora dataset?