I have a doubt Initially you said about number of hops what we could vary, and it is usually 2-3. In pseudo algo, variable K which is referred as depth, is K equivalent to number of hops ?
[Edited] - sorry for the previous one. Was traveling so couldn’t understand the question fully. So essentially, Hops and depths are different things. Hops refer to from how far do we want to take in msg to update current node. Whereas depth, refers to the no. Of layers in Gnn. So essentially at every depth you will have snapshot of the entire graph with certain representation for each node aggregated from k hop neighbors. I hope i was able to explain. Thanks
Oh okay 🤔 maybe I might have missed something. I clearly don’t remember the content of this one now. It would be really helpful, if you could point the error with timestamp and its alternate explanation, it will really benefit anyone who sees it from here on. I’ll also pin the comment to prioritise its visibility. 🙏 thank you
Hey there,
You can watch more Research Paper walkthroughs at - th-cam.com/video/ykClwtoLER8/w-d-xo.html
Cheers!
Great explanation!! Very helpful 🔥
Hey thanks Ratika :)
thank you for such lucid explanation
Thank you :)
Jure Leskovec is such a giant in this field. Such a young guy and doing such a seminal work.
Can’t agree more. 🙇
Great stuff, helps a lot!
Glad to know that. Thanks man!
11:46 I am confused, you made a note about the difference between the psuedo code and the first aggregator function result. And again at 12:50 or so.
Please explain the mini-batch sampling method also from the appendix of this paper.
Can please explain this research paper " Deep learning for technical document classification "
I have a doubt
Initially you said about number of hops what we could vary, and it is usually 2-3. In pseudo algo, variable K which is referred as depth, is K equivalent to number of hops ?
[Edited] - sorry for the previous one. Was traveling so couldn’t understand the question fully. So essentially, Hops and depths are different things. Hops refer to from how far do we want to take in msg to update current node. Whereas depth, refers to the no. Of layers in Gnn. So essentially at every depth you will have snapshot of the entire graph with certain representation for each node aggregated from k hop neighbors. I hope i was able to explain. Thanks
Very nice video 👍
Can you send me slides please
I guess you meant annotated paper. Unfortunately don’t have it now. Space constraints 😅
Your explanation of T2 becoming 0 with sigmoid is incorrect.
Oh okay 🤔 maybe I might have missed something. I clearly don’t remember the content of this one now. It would be really helpful, if you could point the error with timestamp and its alternate explanation, it will really benefit anyone who sees it from here on. I’ll also pin the comment to prioritise its visibility. 🙏 thank you