Logo

ITU AI/ML in 5G Challenge Open House and Roundtable - Shared screen with speaker view
Rakesh Mundlamuri
29:17
Hello everyone, In this meeting will there be a session on ML-PHY channel estimation challenge? also is there someone to answer our questions about the same challenge?
Thomas Basikolo (ITU)
29:51
Yes we have
Rakesh Mundlamuri
30:03
Thanks
Thomas Basikolo (ITU)
30:18
you can write your questions here. They will answer
Rakesh Mundlamuri
30:32
sure, thank you
steven wandale
35:06
Hello, just missed the first which challenge problem is being discussed at this time?
Vishnu Ram
35:29
Franc is discussing https://www.upf.edu/web/wnrg/ai_challenge
steven wandale
36:56
Thanks alot
Rakesh Mundlamuri
40:03
[ML-PHY channel estimation] HI, we have a question about the beamformer and combiners vector generation. From the matlab code used to generate the training data, we can see that for generating the beamformers(Ntrain*Lt) and combiners(Ntrain*Lr), there is a seed rng(1) for the random generation of beamformers and combiners for the number of training symbols.for the receiver to get the beamforming and combining vectors can we use the same seed with Ntrain = number of symbols(20 in the first test set) and can be used for the corresponding symbols for testing?
Rakesh Mundlamuri
42:23
or are we expected to decode the data without the knowledge of these beamformers and combiners? beacuse we cant form the sensing matrix without them
Francesc Wilhelmi
45:30
https://arxiv.org/abs/1910.03510
Francesc Wilhelmi
45:37
https://github.com/fwilhelmi/machine_learning_aware_architecture_wlans
Thomas Basikolo (ITU)
47:08
Now: Barcelona Neural Net-working Center (BNN-UPC) https://bnn.upc.edu/challenge2020
Amjad Iqbal
01:06:02
could we create our own dataset for the said problem by using deep reinforcement learning online method
Muhammad Usman Sheikh
01:09:58
Hi .. my question is there are three categories of problem statements .. Unrestricted, Restricted problems and Problems which are under progress ..... Can I select a problem which is under prgress?or the problems which are Unrestcited are only avialble to the contestants
Reinhard Scholl
01:11:29
Answer for Muhammad Usman: please select "Unrestricted Problems".
Muhammad Usman Sheikh
01:13:04
Thanks
Aldebaro Klautau
01:14:08
Hint for those working in the beam selection problem: you may find useful the new tool Raymobtime_visualizer.py to understand your scenario and eventually “debug”. It is available at: https://github.com/lasseufpa/ITU-Challenge-ML5G-PHY/blob/master/Visualizer/Raymobtime_visualizer.py specially to understand the bias that the scenario imposes (for example: unbalanced dataset)
Khalid Albagami
01:19:14
Where we could find the data ?
Aldebaro Klautau
01:19:45
https://ai5gchallenge.ufpa.br/
Sripada K
01:19:57
[ML phy channel estimation] can we have a walkthrough on details and expectation channel estimation challange?
Vishnu Ram
01:20:52
Visit the website for submitting solutions.
Khalid Albagami
01:21:28
just a moment