GCN on multiple graphs

[mon28]

Hi,

I have a dataset of invoices with each node being each word in the document labeled to the category it belongs to, example - 02/12/2019 is labeled as due_date, $12,000 as total_price, etc.
Here, one invoice page forms a graph. I have 168 such graphs with their adjacency matrix and feature matrix.
How do I prepare the dataset for GCN training using all these graphs?

[tkipf]

For batch-wise training with multiple graphs, please have a look here: tkipf/gcn#4

…

On Thu 4. Jun 2020 at 09:42 Monica Tare ***@***.***> wrote: Hi, I have a dataset of invoices with each node being each word in the document labeled to the category it belongs to, example - 02/12/2019 is labeled as due_date, $12,000 as total_price, etc. Here, one invoice page forms a graph. I have 168 such graphs with their adjacency matrix and feature matrix. How do I prepare the dataset for GCN training using all these graphs? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#61>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABYBYYAPTDTL7ABGRV2EL4TRU5F6BANCNFSM4NSLSFCQ> .

[mon28]

Thanks for the reply. In my case, block diagonal input seems applicable. Do I just concatenate the feature matric one below the other?

[doluong-opt]

@mon28 did you have the answer for this problem ? can you give me the solution for this ?
Thanks