Shulou(Shulou.com)11/24 Report--

Thank you, Mr. Air, a netizen of CTOnews.com, for your clue delivery! CTOnews.com August 18 news, according to Aliyun official account, recently, Ali decision intelligence team developed a large model mathematical tool MindOpt Copilot, which uses the large model as a modeling assistant, receives optimization questions raised by users in natural language, and can automatically model and call a solver to complete the solution.

▲ image source Aliyun official account CTOnews.com also found that the government provided an optimization question with five variables and asked MindOpt Copilot to give the correct answer:

A factory produces five parts A, B, C, D and E, which requires 2 workers, 1 machine, 3 workers, 2 machines, 5 workers, 3 machines, 1 worker, 2 machines, 4 workers and 4 machines.

The profits of each part are 80, 90, 130, 50 and 100 yuan respectively. The factory has a total of 60 workers and 40 machines, with a maximum of 20 for each part.

How to assign production tasks in order to maximize the total profit?

Aliyun said that after the user enters the question, MindOpt Copilot will guide the user to improve the problem description through multiple conversations, then automatically model, program, solve, and finally give the results and analysis of the results.

▲ image source Aliyun official account claims that MindOpt Copilot is based on Ali's self-developed large model, solver and modeling language, and with the help of the underlying general meaning big model, the optimization problem described by users in natural language can be transformed into the optimization model of linear programming and mixed integer linear programming, and the best answer can be obtained. For example, "how to optimize the seating arrangement of restaurants", "how to control risks and maximize corporate profits" or "chickens and rabbits in the same cage".

Aliyun also said that MindOpt Copilot can be used to solve resource allocation and optimization problems in catering, retail, logistics, freight, manufacturing and other scenarios, and will launch Aliyun and provide services in the near future.

Yin Woltao, head of ▲ Tu Yuan Aliyun official account Ali decision intelligence team, said: "generative AI has a lot of knowledge, but at this stage, it is not good at mathematical calculation, so it is unable to generate accurate and easy-to-verify answers to complex optimization problems. In professional fields such as mathematics and logic, the combination of problem-solving AI' and generative AI based on formal modeling language and optimization solver is the preferred way to deal with related problems. "

However, Ali Yun also mentioned that real-world optimization problems are often more complex, and MindOpt Copilot is still exploring industrial-level problem-solving skills. At present, MindOpt Copilot can deal with problems including hundreds of variables or constraints, can also deal with high-dimensional problems that need to be combined with data file input, and can also output mathematical formulas and code, making it convenient for developers and users to make fine adjustments and code development. The R & D team is exploring problems such as overcoming nonlinear functions.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.