Wuyue Cup
![](/competition/wuyue/featured.png)
The Wuyue Cup Quantum Computing Challenge was organized by the China Mobile Ltd. and the Beijing Society of Image and Graphics, and hosted by the Southern University of Science and Technology. The competition attracted nearly 2,000 teams and close to 5,000 participants, including students from approximately 500 various higher education institutions. The challenge was divided into three rounds: preliminary, semi-final, and final. The preliminary round started in October 2023, with the top 20% of teams in each category advancing to the semi-final. The top 12 teams from the semi-final, based on comprehensive ranking, progressed to the final. The final was held on April 27th, 2024 at the Suzhou Jinji Lake International Conference Center. After intense competition with teams from universities such as Sun Yat-sen University, Southeast University, and Tongji University, our team from Tsinghua University emerged as the champion, achieving the first place.
![Gold Medal](/competition/wuyue/certificate_hub2813fd68a25f58b6390a6ee860559aa_2475777_62b52713e34b615125b12bda394869fd.webp)
The challenge required participating teams to model real-world problems into QUBO problems solvable by an Ising machine and to solve them using a quantum computer. Neural networks are the core carriers of artificial intelligence, but training large-scale neural networks faces high costs and long durations. The computational principle of the Ising machine is quantum parallelism, which has great potential for accelerating AI model training. Our team proposed the first Ising learning algorithm to train multilayer feedforward neural networks. Real machine experiments showed that this method could complete the training of the MNIST dataset for handwritten digit recognition in milliseconds, achieving an order of magnitude acceleration compared to the GPU-based backpropagation algorithm. The review experts believe that with the exponential growth of the number of qubits, this method has the potential to reduce the training time of large-scale neural networks, providing a new computational pathway for the era of artificial intelligence.