[nmrg] Graph Neural Networking challenge 2021 - Creating a Scalable Network Digital Twin

[José Suárez-Varela <[e-mail]>]

[Apologies if you receive multiple copies of this email]

Dear Colleagues,

We are glad to announce the "Graph Neural Networking challenge 2021", an 
international competition on Graph Neural Networks applied to computer 
networks, organized as part of the "ITU AI/ML in 5G Challenge".

Challenge description: Creating a Scalable Network Digital Twin

Top-3 teams will have access to the global round of the "ITU AI/ML in 5G 
Challenge <https://aiforgood.itu.int/ai-ml-in-5g-challenge/>". This 
competition attracted over 1,300 participants around the world in the 
last edition, and there were attractive cash prizes for winners.

We provide several resources and tips to easily start working on the 
challenge (See “Main Resources”).

Registration is now open and free of charge for all participants 
(Deadline Aug 31st).

Website: https://bnn.upc.edu/challenge/gnnet2021 
<https://bnn.upc.edu/challenge/gnnet2021>

Registration: https://bnn.upc.edu/challenge/gnnet2021/registration 
<https://bnn.upc.edu/challenge/gnnet2021/registration>

Please, do not hesitate to contact us for questions or 
comments.<mailto:[e-mail]>

=====================================================

[Overview]

Graph Neural Networks (GNN) have produced groundbreaking applications in 
many fields where data is fundamentally structured as graphs (e.g., 
chemistry, physics, biology, recommender systems). In the field of 
computer networks, this new type of neural networks is being rapidly 
adopted for a wide variety of use cases, particularly for those 
involving complex graphs (e.g., performance modeling, routing 
optimization, resource allocation in wireless networks).

The Graph Neural Networking challenge 2021 brings a fundamental 
limitation of existing GNNs: their lack of generalization capability to 
larger graphs. In order to achieve production-ready GNN-based solutions, 
we need models that can be trained in network testbeds of limited size 
(e.g., at the vendor’s networking lab), and then be directly ready to 
operate with guarantees in real customer networks, which are often much 
larger in number of nodes. In this challenge, participants are asked to 
design GNN-based models that can be trained on small network scenarios 
(up to 50 nodes), and after that scale successfully to larger networks 
not seen before, up to 300 nodes. Solutions with better scalability 
properties will be the winners.

[Problem statement]

The goal of this challenge is to create a scalable Network Digital Twin 
based on neural networks, which can accurately estimate QoS performance 
metrics given a network state snapshot. More in detail, solutions must 
predict the per-path mean delay given: (i) a network topology, (ii) a 
traffic matrix, and (iii) a routing configuration.

As a baseline, we provide RouteNet, a GNN model recently proposed to 
estimate end-to-end performance metrics (e.g., delay, jitter, loss) in 
networks. However, RouteNet does not scale well to networks considerably 
larger than those observed during the training phase. As a result, it 
does not perform well when applied to this challenge.

Participants are encouraged to update RouteNet or submit their own 
neural network models.

[Main resources]

- Summary slides (with some tips for participants)

- Baseline GNN model: Open source implementation of RouteNet, including 
a tutorial on how to use it and modify fundamental characteristics of 
the model

- Training/validation datasets

- Python API to easily read and process the datasets

- Mailing list for participants and people interested (Link to 
subscribe:https://mail.bnn.upc.edu/cgi-bin/mailman/listinfo/challenge2021 
<https://mail.bnn.upc.edu/cgi-bin/mailman/listinfo/challenge2021>).

[Timeline]

* Registration open to participants: Deadline Aug 31st
* Evaluation phase: Sep 16th
* Winners (top 3) official announcement: Oct 31st
* Final awards and presentation: Dec 2021

Best regards,

José Suárez-Varela

------------------------

Postdoctoral researcher

Barcelona Neural Networking center (BNN-UPC)

Universitat Politècnica de Catalunya

Contact: [e-mail] <mailto:[e-mail]>