mariokrenn6240 / semnet Goto Github PK

Hello @MarioKrenn6240,

I'm working on my course project and am reading your paper and codes. It's a great idea to consider each node on the graph to be the concept and predict a research trend by doing a link prediction task.

However, when I ran through the python codes, I noticed that line 41 of calc_properties.py is dividing the dot product by the Frobenius norm (np.linalg.norm doc) of the single_net to calculate the cosine similarity between the paper i and j.

This is incorrect because the cosine similarity b/w paper i and j should be irrelevant to the other papers and it should be fixed to

    cos_sim=dot(single_net, single_net)/(
            norm(single_net, axis=1)[:,None]@norm(single_net, axis=0)[None,:]  # (n, n)
    )

The divide here is now element-wise division and the operation is well-defined in numpy.

Could you double-check it?

EDIT: Please ignore the pull request. I realized that if I implement the cosine similarity correctly, there are a bunch of cases that the entire row or column of the single_net are all 0, which results in divide by zero issues and the result becomes a bunch of np.nan.

I noticed the pathnames in various Matlab files (e.g. FromScratch.m) have Windows style back slashes (e.g. '').
(I'm running on a Mac).
Also, in the file CreateFullArticleDataAPS.m:

addpath('C:\Users\IQS-1\Dropbox\MetaNetwork\Indizes\jsonlab')

Hi @MarioKrenn6240
Thanks for making the dataset and code available.

I'm trying to reproduce the dataset using create_full_article_data.py script, but I noticed that the arxiv xml files only include years from 2000 and onwards. Is there a way we get the full article set as used in the pnas paper, from 1919 year, or let me know if I'm missing something.

Thanks.

I enjoyed the paper and am thinking of using some of its ideas in a personal project. Before I start looking at or potentially using some of the code in this repo, I want to be legally in the clear to do so. Can you consider adding a license?

More info is here:

https://docs.github.com/en/free-pro-team@latest/github/creating-cloning-and-archiving-repositories/licensing-a-repository

You're under no obligation to choose a license. However, without a license, the default copyright laws apply, meaning that you retain all rights to your source code and no one may reproduce, distribute, or create derivative works from your work.