Good Evening from Nepal

Yes I am

You're on mute Mr. Reinhard, I can't hear you

Yes

Yes I am I am from Trento, Italy

Calling in from Johannesburg

Good evening from Egypt

from New Delhi, India

Yes. Hi, I am from Colombia.

yes I want to participate (Egypt)

From Toronto

calling from new delhi india

Best wishes for PhD defence!

Best wishes in your defense

Yes

Yes

Is this webminar being recorded?

yes, it is recorded

We take questions here. Please feel free to put your questions here.

Hi . This is Prof. Thanuja. Professor, Karnataka India

hi, this is Abdulaziz Binsaif from Saudi Telecom Company,Saudi Arabia

welcome.

Please Register for the ITU AI/ML in 5G Challenge here: https://www.itu.int/net4/CRM/xreg/web/Registration.aspx?Event=C-00007607

Join our Slack Channel to keep updated: https://join.slack.com/t/itu-challenge/shared_invite/zt-eql00z05-CXelo7_aL0nHGM7xDDvTmA

I've read that when you change the underlying graph, the GNNs eigenfunctions change as well. How can your GNNs generalize to "unseen" networks?

As well I've read that operational backward and forward pass operations take O(n^2), whilst traditional CNNs have O(n). Why do you state that GNNs are faster?

Networks are also based on regulatory & standards guidelines, how do you build that in the GNN ?

Could you please share the slides with us through email?Could you please provide a pytorch version of RouteNet? Now pytorch is widely used.

The Slides can be obtained on the ITU AI/ML in 5G Challenge: https://www.itu.int/en/ITU-T/AI/challenge/2020/Pages/default.aspx

Presentation slides: https://www.itu.int/en/ITU-T/AI/challenge/2020/Documents/Presentation_gnnet_challenge.pdf

I don't know if this is appropriate to ask here. I was trying to run the RouteNet using the main.py code that you provided to see the first test. I get the following problems. I am new to RNN and ML. "... 2020-06-19 14:25:22.119924: I tensorflow/compiler/xla/service/service.cc:176] StreamExecutor device (0): Host, Default Version2020-06-19 14:25:22.129538: F tensorflow/stream_executor/lib/statusor.cc:34] Attempting to fetch value instead of handling error Internal: failed to get device attribute 13 for device 0: CUDA_ERROR_UNKNOWN: unknown error"

I mean new GNN

We can discuss the finer details in the Slack Channel

https://join.slack.com/t/itu-challenge/shared_invite/zt-eql00z05-CXelo7_aL0nHGM7xDDvTmA

Whats the future of GNN in next 10 years especially In health industry

But if you build a kernel filter to act on the "frequency domain", the frequencies, i.e., the laplacian eigenfunction changes.Thus, the learned filter does not apply to the new graph.

Hi, thanks for the presentation! Would you be able to provide RouteNet baseline predictions for training and validation dataset? Thanks

Maybe I missed a part of your explanI was checking the training dataset, but I cannot see a relation between the simulation results and the queueing policy implemented. Each .tar.gz has the results of one policy on all the nodes?

And the MAPE will be measured according to each pair end-to-end delay?

can we have the presentation please

Presentation slides:https://www.itu.int/en/ITU-T/AI/challenge/2020/Documents/Presentation_gnnet_challenge.pdf

from your experience what are the value proposition of 5G ML from an operator prespective please

Maybe I missed a part of your explanation. I was checking the training dataset, but I cannot see a relation between the simulation results and the queueing policy implemented. Each .tar.gz has the results of one policy on all the nodes?

yes please we are very interested from Network prespective

ai5gchallenge@itu.int

Thanks

Join us on the Slack Channelhttps://join.slack.com/t/itu-challenge/shared_invite/zt-eql00z05-CXelo7_aL0nHGM7xDDvTmA

The questions that I missed, we will get back via email or slack. Sorry!