Not long ago, the domestically developed basic industrial engineering software from China once again successfully topped the authoritative international mathematical software evaluation list, and China's independent innovation camp has made continuous breakthroughs in the international basic science and technology arena.
Once upon a time, the term "solver" was associated with the big three: IBM Cplex, Gurobi, and FICO Xpress. Solver is a software for solving mathematical planning problems. It has high technical barriers and is difficult to research and development. There are only a handful of teams capable of research and development in China. As a result, many domestic enterprises, especially some high-tech industries, such as the State Grid, China Southern Power Grid, airlines, etc., used data computing engines are provided by foreign solving companies.
Faced with this situation, Luo Xiaoqu, co-founder and CEO of Shanshu Technology, felt that "we should devote ourselves to developing a 'solver' with independent intellectual property rights." The seemingly simple "solver" is actually very demanding at the technical level. From data collection and calculation to model optimization, the difficulty is the calculation of parameters. Sanshu Solver R & D team spent three years, independent research and development of the first commercial solver platform in China - "Sanshu solver COPT".
Recently, in the latest round of test results of Mittelmann, the most authoritative international third party evaluation in the solver field, Sugishu Technology has ranked first in the performance of simplex linear programming solver in the world. It's worth noting that the solver COPT has been at the top of Mittelmann's list since it debuted in May 2019.
According to the latest ranking, Suginumber continues to rank first, followed by Gurobi, a world-renowned international leader in mathematical programming solvers, and MDOPT, which was released last month by Alida Dharma Institute.
Fir number solver COPT tops the international ranking of authoritative mathematical decision software evaluation
It is understood that the website maintained by Professor Hans Mittelmann of Arizona State University provides evaluation data of multiple commercial and open source mathematical programming solvers, comprehensively evaluating the ability of each software to solve LP, MIP, (MI)QP, (MI)QCQP, (MI)SOCP, NLP and other problem sets. In recent years, his reviews have become a window for potential customers to understand and select solvers, and are regarded by the solver industry as the de facto standard for ranking software performance.
The latest version 2020 provides 40 industrial data sets for open source and business mathematics Planning Optimization solver software evaluation. Evaluation results are sorted by computation speed (solving speed) and completion rate (solving all 40 industrial data sets). For a long time, American and European software occupy the main body and top three of the evaluation rankings. Mathematical programming, also known as mathematical optimization, is the core theory and support of operations research. It is widely used in engineering, economy, finance and other fields to solve the optimal decision of various practical problems, and has great strategic value to modern society. Mathematical programming includes linear programming, integer programming, nonlinear programming and other research directions.
The simplex method of linear programming is regarded as the beginning of modern mathematical programming and operations research, and is also recognized as one of the ten most important algorithms of the 20th century. With the development of information technology, many excellent decision optimization solvers based on simple linear programming have emerged in Europe and the United States and other countries and regions, which can easily solve the linear programming problem of millions of variables. For the warehouse network layout of supply chain, commodity pricing and distribution and replenalization, production and shift scheduling of manufacturing factories, and scheduling optimization of energy grid, Flight recovery and crew scheduling in the aviation industry, as well as transportation planning in the logistics industry and other enterprise optimization decision-making scenarios are of great commercial value.
Based on the massive big data and large-scale business scenarios accumulated by Chinese enterprises, Sanshu Technology uses cutting-edge technologies such as operations research and machine learning to transform practical problems of enterprises into mathematical models. The core product of Sanshu Technology is COPT decision optimization solver, currently offering academic, stand-alone and server versions. Since the release of COPT version 1.0 in May 2019, the solver has been continuously improved and is now able to successfully solve linear programming problems with 50 million to hundreds of millions of variables in practice, according to Sugisuo.
For mature commercial solvers, the technical research and development difficulty of integer programming (MIP) solvers is much higher than that of linear programming, which requires a highly complex mathematical system to be engineered through millions of codes, and involves more sophisticated optimization mathematical principles. Research and development personnel need to have excellent computer programming ability and profound mathematical and optimization algorithm theory, so the integer programming module is also the part of the number solver team in recent years.
At present, COPT integer programming module (MIP) has tended to mature, after internal testing, the performance is about 20% better than the best MIP open source solver SCIP, not yet open to the whole society for trial, but has been applied to a number of head enterprises, with the ability to quickly and stably deal with various integer programming problems in the industry scene. For example, State Grid California Research Institute combined optimization of water, fire and electricity units, Budweiser intelligent supply chain management system, Xiaomi Omni-channel intelligent cargo distribution system, Comac civil aircraft industry knowledge graph optimization algorithm engine deployment, China Southern Airlines 2020 annual engine optimization project, Holliyou supply chain management digital construction project, etc.
Ge Dongdong, co-founder of Sanshu Technology and project leader of COPT decision optimization solver, said that as China has become the world's largest industrial manufacturing country, the largest network retail country, the largest trade country, the world's second largest economy, China's business scene and scale has been no less than Europe and the United States, This provides a rich scenario for domestic industrial engineering software research and development represented by mathematical programming solver, as well as reaching or even catching up with the international top commercial software level, and creates a historic development opportunity.
COPT solver has been widely deployed in the smart supply chain and Industry 4.0 intelligent systems of Suginumber Technology in the past two years. In addition to ranking the top of Mittelmann linear programming test since May 2019, COPT solver modules have been widely used to solve practical large-scale optimization business problems and national important industries, especially in linear programming has reached the level of fully replaceable foreign software. The fir solver COPT currently supports all major operating systems including Windows, Linux and MacOS (all 64-bit), It also provides interfaces to major computing languages such as Python, PuLP, Pyomo, C, C++, C#, Java, AMPL and GAMS, as well as support for the ARM64 platform.
Luo Xiaoqu told reporters: "COPT solver is the core product and underlying technology of Sanshu Technology, driving the upper industry solutions and services, Sanshu technology for retail, logistics and manufacturing industries to provide intelligent forecasting and promotion optimization, intelligent pricing, intelligent replenishment, intelligent planning and production, transportation optimization, intelligent storage, intelligent location and other general solutions, To help enterprises in the Chinese market achieve optimization and transformation from data to decision making."
Sugigu, some say, is an atypical AI company. It does not engage in any of the voice, image and NLP chain, nor does it take machine learning and deep learning as the only core support. Sanshu Technology stands at the intersection of machine learning and operation research, and does the business of modeling, solving and providing the best decision after data rule mining.
In this regard, Luo Xiaokui bluntly said: "From the beginning to the end, Shanshu Technology has been doing two things, the first thing is' solver development ', the second thing is' create intelligent decision wisdom chain '." At present, the era of big data has brought about digital "butterfly change" in almost all fields, and Shandata Technology will continue to mine and integrate big data information, drive enterprises to make innovative decisions, and give greater play to the value of "data storm".
The year 2020 is the first year of the plan for the 14th Five-Year Plan. It also meets the overall strategic agenda for the great rejuvenation of the Chinese nation. Faced with the international environment full of uncertainties, the new generation of domestic industrial engineering software enterprises represented by Sanshu Technology will unswervingly build independent and controllable domestic basic software, and inject China's scientific and technological responsibility for the great changes in the world unseen in a century.
