Position: Home > News > News > Content

​Tsinghua University student triumphs in prestigious programming contest

Jun 29, 2023 08:39

Wang Jiayi, a 2020 graduate student from the Department of Computer Science and Technology, Tsinghua University, has won the ACM SIGMOD Programming Contest 2023, a prestigious international competition in the field of data management, under the supervision of professor Li Guoliang. This is the first time that Tsinghua University has won the championship in this event.

The award ceremony of the ACM SIGMOD Programming Contest 2023 was held on June 20 in Bellevue, Washington. The ACM SIGMOD Programming Contest is an annual event held in conjunction with the ACM SIGMOD International Conference on Data Management, one of the best academic conferences in the database field. The contest challenges students to solve a real-world problem in data management using innovative and efficient solutions.

The task of this year’s contest was to build and query K-nearest neighbor graphs on massive vector data sets. A K-nearest neighbor graph is a data structure that connects each vector in a data set with its K closest vectors. Such graphs can be used as indexes to improve the performance and accuracy of approximate nearest neighbor queries, widely used in applications such as information retrieval, recommendation systems, and natural language processing.

The test data set in the contest was 10 million Bing queries encoded by Microsoft’s large-scale natural language representation model Turing v5 into 100-dimensional vectors. The contestants were asked to build a K-nearest neighbor graph with high recall rate (the percentage of true nearest neighbors that are found) while supporting fast concurrent queries.

Wang Jiayi proposed a new solution based on the classic nearest neighbor descent algorithm NN-Descent. His solution achieved a recall rate of 98.7% close to the optimal within a limited construction time, outperforming all other contestants.

A total of 28 teams participated in this competition, including teams from University of Michigan, Technical University of Munich, UC Riverside, National University of Singapore, Hong Kong University of Science and Technology, Fudan University, Shanghai Jiaotong University, among other world-renowned universities.