On December 30, Beijing time, COPT optimization solver once again topped the list of the international authoritative third party evaluation Mittelmann platform linear programming simplex method, continuing the algorithm module of the first place in the world. Compared with the September version, the optimization solver COPT algorithm performance has achieved 1.9 times of improvement. At the same time, in the test list of linear programming interior point method, it also achieved a comprehensive speed increase, maintaining the global runner-up position after Gurobi. It also surpassed Gurobi for the first time on the test list of linear programming large-scale network problems.
Mittelmann evaluation platform is personally maintained and operated by Professor Hans Mittelmann from Arizona State University. The benchmark problem library used for evaluation is derived from the classic practical problems in the industry. Its evaluation has been unanimously recognized by the solver industry and is known as the "lighthouse" for judging the performance of solvers. The figure shows the evaluation results of Mittelmann linear programming simplex method. The number "1" means that COPT has the fastest solution speed and serves as the benchmark time, and the number "40" means that COPT has successfully solved all 40 linear programming examples. This COPT linear programming simplex method not only broke the world record, but also proved once again that the overall level of domestic solver has reached the world first-class.
Figure 1: COPT continues to top the list of Mittelmann linear programming simplex methods
Figure 2: COPT leads the list of linear programming interior point methods globally
Figure 3: COPT tops the list of linear Programming large-scale network problems for the first time
It is worth emphasizing that the fir number optimization solver COPT is the first industrial solver independently developed in China, and the only comprehensive solver in China that has the capability of solving both large-scale linear programming (simplex method and interior point method) and mixed integer programming. Since its release in May 2019 as the first commercial-grade solver in China, it has participated in the Mittelmann Test platform competition and won the first place in the list, bringing China's solver to the world stage. In the year and a half since its birth, especially in 2020, after several fierce competitions, it has always maintained a strong dominant position in the first place for most of the time.
Since its birth, COPT has been open to the whole society, users can download free use. Over the past year and a half, more than 400 education, enterprise and government users have applied for licenses, including more than 50 overseas users.
Ge Dongdong, Co-founder &CSO of Sugigu Technology, said, "This improvement is largely due to our research on some new technologies, such as the successful parallelization of the dual simplex of linear programming. As far as we know, this function has been Gurobi, Cplex and other solver giants headache, do not do very good difficulty. In addition, Crossover algorithm, for example, belongs to the direction that has been relatively quiet in the international field in recent years. Our professional team achieved some new breakthroughs in thinking this time and successfully applied it into the solver. "
At present, many modules of integer optimization, linear optimization, and nonlinear optimization of COPT have been applied in retail, manufacturing, energy, finance, aviation, military and other industries, serving customers including Haier, Xiaomi, Comac, State Grid, China Southern Airlines, Six countries Chemical, Holliyou and other leading enterprises in more than 20 industries. Help it solve optimization problems in supply chain management, production planning, production scheduling, process optimization, transportation, scheduling optimization, maintenance planning, marketing strategy and other scenarios.
