Transfer learning for detecting failures of optical access network
June 16-30th 2021
Huawei Paris Research Center is thrilled to announce the European University Challenge 2021 in France, Spain and Portugal
Huawei European University Challenge is an annual competition across Europe. As an industry leader in the information and communications technology infrastructure and smart devices industry, Huawei is eager to work with the next generation of innovative thinkers.

Optical networks constitute a big part of today's internet infrastructure and are key to bring the information from servers to your home. From time-series data collected from the optical lines in a city, the challenge will be to predict whether the optical link will collapse in a near future.

We will also want your solution to be able to generalize across multiple cities: you will be given two asymmetrical sets of time-series from two different optical networks gathered in two different cities. We obtained a lot of data from the "source" city and much less for the "target" city. The goal of the challenge is to build a transfer learning approach in order to fully exploit the available data to propose the best possible prediction for the optical network of the target city.
Invited to participate
We are looking for undergraduate, Master's, and PhD students currently registered with an institution in France, Spain or Portugal and studying engineering, mathematics, computing science or related fields. Students who graduated up to 2 years ago are also eligible
Why participate
Compete with fellow participants in the competitive phase for money prizes in an exciting data challenge format
Exchange your solutions with your fellow participants in the open phase and watch how the blended score of the community improves
Learn data science from each other, especially during the open phase
Get hired
Make yourself visible and demonstrate your data science prowess to companies who are looking for fresh talents
Prize pool
The final ranking is determined by the accuracy of the model on the target city, measured on a private data set
7000 €
First place
4000 €
Second place
2000 €
Third place
Competition rules summary
See detailed competition rules for more details.

Participants will sign up individually at this site and form teams of one to four persons for the challenge. Teams will have between June 16th and June 27th to submit their solutions, maximum one per day
Members are eligible to this competition if they are registered currently or were registered and graduated not earlier than 2019 in a higher education institute in France, Spain, or Portugal
All models will be trained on the same cloud server allowing 16 CPUs (with shared memory of 128GB RAM)
Show more
Eligible Members may sign up to the competition, to become a Participant. A Participant will be initially assigned to his/her special individual team. If Participants wish, they may form Teams of maximum four members using the "my teams" view in their account, before making any submission. Participants can leave and join Teams freely, under the limit of four members, as long as they, or the Team they would like to join, has not yet made a submission. Once a Team (individual or multiple member) makes a Submission, it becomes immutable. Any Participant can leave a Team, which has not yet made a Submission, on his/her own accord. Joining a Team will require the approval of all existing Team Members.
Teams will be given a total of 20 machine hours per cross validation fold. Submissions of a given Team will be ordered by Submission timestamp. We will make an attempt to train all Submissions, but starting from (and including) the first Submission that makes the Team's total training time exceed 20 hours, all Submissions will be disqualified from the competition. Testing time will not count towards the limit. Training time will be displayed on the leaderboard for all Submissions, rounded to second. If a Submission raises an exception, its training time will not count towards the total.
There is a timeout of 24 hours between Submissions. The only exception is when a Submission cannot be trained due to an error. In this case the Team can resubmit a Submission under the same name immediately.
Submissions submitted after the end of the competition will not qualify for prizes.
The public leaderboard will display validation scores running a cross-validation. The official scores will be calculated on the hidden test set and will be published after the closing of the competition. We will rank Submissions according to their Average Precision score.
The organizers will do their best so that the provided backend runs flawlessly. We will communicate with participants in case of concerns and will try to resolve all issues, but we reserve the right to make unilateral decisions in specific cases, not covered by this set of minimal rules.
The organizers reserve the right to disqualify any participant found to violate the fair competitive spirit of the Data Challenge. Possible reasons, without being exhaustive, are multiple accounts, attempts to access the test data, etc.
Participants retain copyright on their submitted code and grant reuse under BSD 3-Clause License.
Participants accept these rules automatically when signing up to the "Transfer learning for detecting failures of optical access network" Data Challenge. Participation to the Challenge implies full acceptance without reservation of the present terms, the Challenge Rules, and the Platform Terms and Conditions.

Persona data are subject to the Privacy Policy.
May 27th
Registration Opens
June 16th 6 pm
Opening Ceremony
June 27th
Project Submission Deadline
June 30th
Closing Ceremony
Opening ceremony replay
What is the sign-up process?
First you need to create an account on the RAMP platform. Your sign-up request will be approved after you send your proof of student status and expected date of graduation to Jiao Li (li.jiao@huawei.com) with a subject line "Huawei RAMP sign-up".

Then, shortly before the start of the data challenge, you will need to join the corresponding event on the platform. Starting June 16th you can make submissions either individually or as a team.
How can I participate with a team?
All team members need to have an account on the RAMP platform and join the corresponding event. Then you can create a team via the corresponding interface.
Note that you can only make submissions either individually or in a team, but not both. So if you made submissions on your own, you will not be able to join a team. And similarly, if your team has made submissions, you will not be able to exit the team to make individual submissions.
I have a question or something is not working on the RAMP platform.
Please ask questions in the following Slack channels,

- #general: for general questions, including for instance the sign-up process
- #installation-issues: for questions regarding how to use starting kit
- #server-submissions: for any questions about, or issues with the RAMP platform
Data Challenge Overview
Challenge winners
Mathis Federico
1st place
Yizhou Chen
2nd place
3rd place

Haolin Pan, Kunhao Zheng, Yinming Qin

Team C
4th place

Kwate Dassi, Brown Ebouky, Fandio Njylla Esdras Emma, Armelrandy
5th place
6th place
About RAMP
RAMP was originally developed as a tool for data scientists to efficiently and collaboratively solve the data analytics aspect of high-impact domain science problems. Since then, RAMP has undergone many iterations and has been used in teaching and benchmarking machine learning algorithms as well.
No cookies are used for mere visitors of this web site, but access to content is restricted to register Members only.
This site uses certain cookies to remember Members who accepted the conditions when registering with the Web Site and log in as further described here.
About how to set cookies click here.
To see our Privacy Policy click here.
Ok, don't show again