ssgood / mmgl Goto Github PK

I would like to reproduction the results of sMci vs. pMci in your paper. But in the run the code reports an error. I have set PMCI_selection to True in the data preprocessing stage, can I refer to your setting?
Thank you！

Greetings Sir, I am getting the following error related to feature selection function:

selector = RFE(estimator, num_features_selected, step=200, verbose=1)
TypeError: init() takes 2 positional arguments but 3 positional arguments (and 2 keyword-only arguments) were given

Hey! Great work. I was analyzing your paper, and couldn't figure out if you had reported results from inductive learning experiments. Are Table III results from inductive experiments?

Best.

Hello, I am very interested in your work and started trying to reproduce this work. I'm validating based on the TADPOLE challenge data first, but based on your script, the training and testing results are much lower than what is reported in your paper. In addition, I see in your Readme file that the configuration based on that sh file can reproduce the results in the paper, but I tried it and the performance was poor. Therefore, I would like to ask if there are other parameters that need to be set? Looking forward to your reply, thanks!

运行 feat_256_1 = feature_selection(features, label, train_index, 256)函数时
报__init__() takes 2 positional arguments but 3 positional arguments (and 2 keyword-only arguments) were given
我查到的时参数不匹配，但由于调用的是封装函数，我想的可能是sklearn版本不对，请问作者的sklearn版本是多少？

为什么拿一折的train_idx进行特征选择后的256维特征，最后拿去进行5折交叉实验呢？不应该是每折实验都进行一次特征选择吗？

Hello, the MMGL model you proposed is very enlightening. I would like to ask you some detailed questions about the comparative experiment in your article. When I used the LGL model to classify the ABIDE dataset, the average accuracy of the classification results obtained was close to the 86.40 marked in your article, but the variance of the results was large, nearly 7, far greater than the 1.63 marked in your article. Because the author of the original LGL model did not conduct classification experiments on ABIDE, I would like to seek your help. May I ask if it is convenient for you to provide some implementation details about the LGL model for ABIDE classification, such as the need to use recursive feature elimination to reduce the dimensionality of input features? How many graph convolutional layers were used? What is the final MLP dimension size?

你好，我运行了attn_vis.py想可视化注意力矩阵，但是它提示我找不到result中的文件，源码中也没有这个文件夹，我想知道这个问题怎么解决？

恭喜！非常精彩的工作！

这里探讨个问题：

代码里似乎是将 train set 和 test set 统一输入构建 graph，train phase 中 test set 不参与 bp。

如果是这样，那么论文中的

“ the unseen patient u is first added to the existing population graph by MARL with AGL. ... It should be noting that the test patients are not kept in the graph in the testing phase. ”

该怎么理解呢？

期待回复，谢谢！

请问作者在使用GNN实现卷积的时候，是否是通过节点特征矩阵*邻接矩阵的值（邻接矩阵的值是由节点特征的相似性计算得到的）来进行消息传递和更新的，而不是与邻接节点的特征进行直接的交互？我看代码中是这样的，不确定是否理解正确。

Hi guys, we are trying to reproduce your experimental content, can we get the requirements.txt to reconstruct the experimental environment?
Since the dgl version has changed a lot, we can't reproduce your experimental environment by using some newer or older versions.

thanks

when i try to run the command block:
"files = ['rois_ho']
num_subjects = 871
filemapping = {'func_preproc': 'func_preproc.nii.gz',
'rois_ho': 'rois_ho.1D'}

if not os.path.exists(data_folder): os.makedirs(data_folder)
shutil.copyfile('./subject_IDs.txt', os.path.join(data_folder, 'subject_IDs.txt'))"
in /ABIDE/data_processing.ipynb
it reports:FileNotFoundError: [Errno 2] No such file or directory: './subject_IDs.txt'

but i have already downloaded all the ABIDE datasets files, what should i do to run the process
Thanks!

Hi guys, we are trying to reproduce your experimental content, can we get the requirements.txt to reconstruct the experimental environment?
Since the dgl version has changed a lot, we can't reproduce your experimental environment by using some newer or older versions.

thanks