Hi, Jacob! How are you? I'm an undergraduate student from Brazil and I'm working exactly on the problem TGN tries to address (Representation Learning for Event Graphs). Unfortunately, I didn't have the chance to really dig deeper into the paper, but it looks a lot like a mix between a standard Message Passing Neural Network and some adaptations to adapt it to Dynamic Graphs, which seems very powerful from the Representation Learning point of view. Still, in the video, you talk about the fact this work uses timestamped graph events, which may be limiting. Could you elaborate on this thought? Thanks for the amazing work!
Hi Jacob. I'm trying to solve a problem for my internship where I somehow have to translate the edges as not being instanteneous... For example nodes are warehouses and edges are available roads in between, but I need to have something more than "instant" time. i was thinking of something like a feature on the edges which would translate to real time of a ETA of that trip , or something like that . I can't seem to find any papers that deal with that, or with node regression on dynamic graphs of any sorts..
@@deltamico TGN paper from twitter team for structure I guess and regarding regression I guess we could bin the values and do multiclass classification..
Thanks for the feedback! The pixelation is one of the downsides of using my phone as my main camera. I'm making a few upgrades to my video production setup for future videos, so stay tuned!
I would love to see a new update video on the topic in 2023. I am using GNNs for my PhD
Really well-presented video and a nice explanation about TGN. Thanks for that!
The cool guy appeared from nowhere, dropped top-quality content, and just disappeared.
Great work!
Hi, Jacob! How are you? I'm an undergraduate student from Brazil and I'm working exactly on the problem TGN tries to address (Representation Learning for Event Graphs). Unfortunately, I didn't have the chance to really dig deeper into the paper, but it looks a lot like a mix between a standard Message Passing Neural Network and some adaptations to adapt it to Dynamic Graphs, which seems very powerful from the Representation Learning point of view. Still, in the video, you talk about the fact this work uses timestamped graph events, which may be limiting. Could you elaborate on this thought? Thanks for the amazing work!
Hello Jacob, is there a model to use TGNN in keras or tensorflow?
Hi Jacob. I'm trying to solve a problem for my internship where I somehow have to translate the edges as not being instanteneous... For example nodes are warehouses and edges are available roads in between, but I need to have something more than "instant" time. i was thinking of something like a feature on the edges which would translate to real time of a ETA of that trip , or something like that . I can't seem to find any papers that deal with that, or with node regression on dynamic graphs of any sorts..
Have you found some way?
@@deltamico TGN paper from twitter team for structure I guess and regarding regression I guess we could bin the values and do multiclass classification..
Nice!
Hi Jacob ,can you model COVID-19 by GCNs and implement for it. Thanks for the amazing work
Multilayer Graph NN are awesome
Hi Jacob, check the first part of the video: it starts very pixelated/grainy. Otherwise excellent content as usual.
Thanks for the feedback! The pixelation is one of the downsides of using my phone as my main camera. I'm making a few upgrades to my video production setup for future videos, so stay tuned!
Everything changed once you ceased sending in videos:( Lord, return; your people are waiting for you!