net-titech / gnn-models

Graph Neural Network Models in Pytorch
MIT License
3 stars 1 forks source link

Init paper collection #4

Open gear opened 5 years ago

gear commented 5 years ago

Please initiate the paper collection!

Kaushalya commented 5 years ago

I propose to use Github issues to keep track of the paper list. For example, check issue #5. Once we read a paper we update the issue with a new comment containing a brief summary. If we need to have a detailed summary with MathJax, create a new markdown file in the repo and link it to the issue. @gear @zepx What do you guys think?

gear commented 5 years ago

@Kaushalya that is a great idea. I think adding a tag "paper" for all papers would help with the filtering too.

gear commented 5 years ago

Btw, should a tag name "New GNN"? Or should you split it into two tags? "New" and "GNN"?

zepx commented 5 years ago

Btw, should a tag name "New GNN"? Or should you split it into two tags? "New" and "GNN"?

Is the word New necessary? If we're trying to classify papers that we've not read, maybe we should have tags separated for this.

I propose to use Github issues to keep track of the paper list. For example, check issue #5. Once we read a paper we update the issue with a new comment containing a brief summary. If we need to have a detailed summary with MathJax, create a new markdown file in the repo and link it to the issue. @gear @zepx What do you guys think?

Highly agree with this. This also helps us keep track who's the one responsible for reading and understanding that particular paper. It's also easy to ask that person to explain whenever necessary; helps with delegation of work too.

Kaushalya commented 5 years ago

Is the word New necessary? If we're trying to classify papers that we've not read, maybe we should have tags separated for this.

"New" is not necessary. Added that as a test. Better to have an additional tag to mark papers that have been read. I'll start adding papers this week.

gear commented 5 years ago

"New" means "new architecture", right? Since many models are enhancement (SGC, FastGCN, etc), and some models are "new", don't we need this tag?