Replication of approach

[Gautamshahi]

Hello,

If I would like to replicate the model, especially for the ampligraph, how can I do it?

I can not follow the flow of the code.

Can you please share some insights on it?

[shawngraham]

Hi,

The version we used of Ampligraph was 1.4.0, which is now out of date. You can run our notebook in Google Colab though by downgrading the tensorflow version first.

So run this: https://github.com/XLabCU/organigram/blob/main/link-prediction/ampligraph_modelling/ampligraph_modelling.ipynb in google colab and then specify ampligraph 1.4.0 in the second cell:

!pip install ampligraph==1.4.0

Have our knowledge-graph.csv file in the same folder, then proceed through the cells.

[Gautamshahi]

Hello, Thank you for your reply. Yes, I moved to the Ampligraph 2. The problem is that while running query_topn, sometimes the best relation predicted comes at 1 or 4, or 10. So, to compute the error between the actual value and the predicted value how would pick the n? Another point: Did you computerize the error analysis of the predicted value apart from MRR? Since the difference in trade value in high. Thank you for your time. Regards,

…

On Thu, 29 Feb 2024 at 15:39, Shawn Graham ***@***.***> wrote: Hi, The version we used of Ampligraph was 1.4.0, which is now out of date. You can run our notebook in Google Colab though by downgrading the tensorflow version first. So run this: https://github.com/XLabCU/organigram/blob/main/link-prediction/ampligraph_modelling/ampligraph_modelling.ipynb in google colab and then specify ampligraph 1.4.0 in the second cell: !pip install ampligraph==1.4.0 Have our knowledge-graph.csv file in the same folder, then proceed through the cells. — Reply to this email directly, view it on GitHub <#1 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AA6IKFSUQ5EYYQZCNFMJPMDYV46QNAVCNFSM6AAAAABD37YT6GVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTSNZRGI4DIMZTGI> . You are receiving this because you authored the thread.Message ID: ***@***.***>

-- Regards, Gautam Kishore Shahi