This Challenge covers three problems. Each problem will be released on a different date and have its own reward plan. Contestants can choose to work on one or more of the problems.
(The Organizing Committee has decided that Problem B and Problem C for this Challenge will be released at later dates than planned, considering that there are multiple holidays worldwide at the beginning of this year, and that many contestants working on Problem A have also expressed interest in Problems B and C. We hope that this postponement will give contestants more time to take a break, think, and tackle the problems. For more information about the new dates of Problem B and Problem C, see the problem timelines below.)
Problem AData Center Network
Problem BSpatial Estimation
Problem CMulti-Parameter Wireless Network
Problem AData Center Network
Problem A introduction:
Network transmission is one of the key part of distributed storage systems. Network transmission performance directly affects the comprehensive performance of storage systems. Proper network transmission protocols can effectively reduce network latency and tail latency, improve bandwidth utilization, and reduce network jitter. Currently, there are multiple interconnected physical links between most nodes in cloud data centers. Traditional single-path transmission protocols cannot fully utilize the advantages of multi-path transmission. Therefore, how to make full use of the multiple existing physical links to perform multipath transmission and implement load balancing between the multiple links becomes a key to reducing a network transmission latency. Then, implementing an appropriate multipath load balancing algorithm presents a significant challenge.In order to solve these challenges, the contestants need to design a transmission protocol that can implement multipath load balancing according to the dynamic changes of network status. The following objectives are achieved, including the lowest average data flow transmission latency, data flow tail latency, latency jitter, minimized network packet loss, and maximized network bandwidth utilization.
Problem BSpatial Estimation and Separation
Problem B introduction:
In the field of information technology, the issue of generating and jointly processing multi-dimensional data is becoming increasingly common with the enhancement of hardware deployment capabilities. There exists one category of classic issues in this area, known as subspace analysis. It includes analysis on the characteristics of data space to obtain essential information on low dimensions and also includes subsequent processing (such as enhancement or suppression) of this information. This topic focuses on an important issue in subspace analysis: accurate separation and estimation of target linear subspaces in a full-dimensional matrix space. At the core of this interactive topic is to explore the design of solutions for separating and estimating target linear subspaces. As the target linear subspace may be obscured by interfering linear subspaces (that is, other components in the space), participants need to accurately separate and estimate the target subspace with a black box system (linear-nonlinear-linear system). The difference between target and interfering linear subspaces is their respective spatial characteristics. The input design of the black box system can help accelerate the subspace separation and estimation. Note that this topic requires participants to design solutions within a limited complexity (program running time).
Problem CMulti-Parameter Wireless Network Optimization Based on Coverage Simulation
Problem C introduction:
The signal strength received by mobile users is a key indicator of network service quality in wireless networks. This signal strength is influenced by environmental factors, such as building occlusion, and network parameters, including the tilt of panel antennas. To ensure stable network services in a dynamic environment, network operators must continuously adjust these parameters. Typically, they provide network simulation models that capture the quantitative relationship between the parameters and signal strength, enabling them to calculate optimal antenna configurations through model interactions. However, this interaction can be computationally expensive due to the complexity of the models. In this contest, participants are expected to design efficient and intelligent algorithms for network optimization. These algorithms should obtain fitness values (representing average signal strength) from a given black-box simulator but with limited simulations due to their computational complexity. Specifically, multiple candidate parameter configurations are provided for each antenna in the wireless network. The optimizer should explore these candidates to identify the best configuration that maximizes fitness value attainment. Due to an enormous search space and time constraints, brute-force search is not feasible for solving this problem. Contestants can employ various general methodologies like exploration heuristics design, simulation surrogates or search space pruning techniques to address this challenge effectively. Finally, higher scores will be awarded to algorithms that offer better optimal solutions within the specified time limit.
Mathematics and algorithm enthusiasts from around the world
Team competition + Coaching
Team formation: Each team may have at most four people, including a team leader.
Team leader: The team leader creates the team. As the team's contact person for the organizers, the team leader shall receive any prize money on behalf of their team.
Coach: This Challenge provides a coach reward plan, which encourages participating teams to invite domain-specific teachers to act as coaches and help the contestants improve their problem-solving skills during the contest. A coach cannot be a contestant on the team. Failure to provide coach information during registration is deemed as a decision not to participate in the coach reward plan.
Step 1: The team leader scans the registration QR code of the problem to open the registration page, and fills out the required information to complete registration. Once registration is finished, a team QR code is generated, which the team leader shares with their team members.
Step 2: Each team member scans the team QR code to fill out the required information on the registration page and complete registration.
(1) A registration QR code is provided for each problem. Any contestant who intends to work on multiple problems should register for each problem separately.
(2) To register for a problem, team members should scan the team QR code shared by their team leader, rather than the registration QR code of the problem.
(3) During registration, contestants who already have a Kattis account should use the e-mail address they used when creating their Kattis account. Once the Challenge begins, these contestants can log in to Kattis with their own account and password.
(4) Contestants without a Kattis account will receive one, along with a password, after registration;
During this Challenge, there will be two rounds of winner selection and prizes for each problem. In Round 1, winners are selected based on their code. In Round 2, winners are selected based on their code and articles. This section provides information about submission requirements, winner selection rules, and the prizes for Round 1 and Round 2. For more information about the timeline of each problem, see the "About the Problems" section.
(1) Given that writing an article takes time, the last day of article submission is one week later than the last day of code submission. Please keep the deadline for article submission in mind and make necessary preparations.
(2) If submissions are found to be too similar during the review, the relevant teams may be disqualified.
(3) Applicable time zone for all problems: UTC+08:00.
(1) Contestants log in to Kattis to submit code for the problem before the deadline.
(2) The review team for the problem performs a review based on the rank lists on Kattis, and announces the list of winners in Round 1 on this website. See the “About the Problems” section – Problem timeline – Round 1 results released.
(3) Prizes for Round 1 (All amounts post-tax):
Top 1–3 teams:
EUR6,000 per team
Top 4–10 teams:
EUR3,000 per team
Top 11–20 teams:
EUR1,000 per team
Top 100 teams and their coaches:
One T-shirt per person
(1) After Round 1 results are announced, each of the top 40 teams must write an article using the template and submit it in PDF format to challenge4IMC@huawei.com before the deadline. Teams that fail to submit an article before the deadline will be disqualified from consideration during Round 2. Click here to download the template .
(2) The review teams of problems read and score the articles based on their abstract, assumptions and symbols, analysis of the problem, model building, model solving, model summarizing, test results description, references and appendices, and structure and typesetting.
(3) In Round 2, the total score is 100 points, with 60 allocated to the code submitted in Round 1, and 40 allocated to the article submitted in Round 2.
(4) Prizes for Round 2 (All amounts post-tax):
Top 1 team:
EUR4,000 for the team,
EUR1,500 for the coach
Top 2–4 teams:
EUR2,000 per team,
EUR1,000 per coach
Top 5–10 teams:
EUR1,000 per team,
EUR800 per coach
Top 11–20 teams:
EUR800 per team,
EUR500 per coach
Top 21–40 teams and coaches:
One HUAWEI FreeBuds
(5) Coach reward plan for the coaches of outstanding teams in Round 2:
Within two weeks after Round 2 results are announced, coaches of the top 40 teams should submit a summary of their coaching experience using the template to challenge4IMC@huawei.com. Failure to provide a summary of coaching experience on time is deemed as a decision not to participate in the coach reward plan.
(6) The Round 2 results will be released on this website. See the “About the Problems” section – Problem timeline – Round 2 results released.
In addition to the above prizes, the top 100 teams will also be given opportunities to meet Fields medalists and problem authors face to face, and visit Huawei's campuses around the world.