The 2022 GraphChallenge, one of the most influential international competitions in graph computing, released its results on August 15. Following their first victory in the GraphChallenge for China in 2021, two of the graph computing teams of the Service Computing Technology and System Lab, School of Computer Science and Technology, HUST, have won the world champion for the second time.

GraphChallenge, which is one of the most influential international competitions in the field of graph computing, has been held six times. The GraphChallenge, which includes three courses: Static Graph Challenge, Streaming Graph Challenge, and Sparse Deep Neural Network Graph Challenge, attracted many well-known companies and universities from around the world, including NVIDIA, Xilinx, Huawei, Lawrence Livermore National Laboratory, Carnegie Mellon University, Washington State University and so on. In the past years, the competition was dominated by well-known American universities and enterprises. In 2021, however, the graph computing team of  HUST won the championship for China for the first time. The team achieved yet another feat this year, winning the two champions for the second time in a row.

The two winning entries were completed by senior undergraduate student Sun Yufei and doctoral student Wang Qinggang, doctoral student Xu Shaoxian and master student Wu Minkang, and supervised by Associate Professor Zheng Long, Professor Shao Zhiyuan, Professor Liao Xiaofei, and Professor Jin Hai. According to Wang Qinggang, the team members started preparing for the competition at the beginning of this year and decided to focus on the Sparse Deep Neural Network Graph Challenge. Every week, they discussed with Huang Yu, Yao Pengcheng, Ye Xiangyu, and other teammates, fully analyzed the advantages and disadvantages of the 2021 champion scheme and sought the optimized solution. The two entries adopt different schemes from different perspectives to improve the performance of sparse neural network inference. Together, the team’s two entries won the world championships with performance improvements of 8.12 times and 6.37 times over the same track champion last year.

According to the winning team, a graph is a kind of flexible data structure composed of vertices and edges, which can be used for modeling based on the relations in the real world. Therefore, graphs are necessary for many areas in real life, including financial transaction graphs, state grid graphs, virus spreading graphs, and traffic road graphs. In order to mine useful information from huge graph data, graph computing technology emerged. It is now widely used in many fields, such as financial fraud detection, power grid troubleshooting, virus transmission tracking, travel road planning, and military intelligence analysis. At present, graph computing has become a new area of rivalry for both domestic and foreign technology giants. However, the sparsity and irregularity of real graphs make it more difficult to efficiently process and analyze large-scale graph data. In 2017, the IEEE, MIT, and Amazon jointly created the GraphChallenge, an international competition that calls on global researchers to develop new solutions to improve the processing efficiency of graphs and sparse data from real life.

Zheng said that in 2021, the graph computing team won the championship for the first time in the competition, ending the dominance of well-known research institutions in the United States, led by NIDIA and Lawrence Livermore National Laboratory. In 2022, the team won the double championships for a second time, which is also of great significance. The team will continue to deepen its exploration in the field of graph computing and maintain its lead in this field with more excellent results.

Written by:Meng Ziyun

Edited by:Shi Can, Peng Yumeng

Source: School of Computer Science and Technology