Thanks for the brief overview! As you mentioned that there are some high-level APIs to quickly model scheduling or VRP-like problems, I was wondering if these API's can be used to model a problem that contains both routing and scheduling. An example of such can be home service assignment problem where we need to optimize the total cost to hire servers that serve multiple customers and also schedule their routes such that the customers are served in their preferred time duration. Your thoughts on this would be highly appreciated!
I have to admit Im not a programmer so Im quite laic but I was always wondering why I need these programming tools if I can write my problem by an algebraic modeling language for example AMPL and just "give" my model to a solver. What are the advantages of formulate my problem in C++ or any other programming language?
Because most programming languages are open source and can do what commercial packages can do. Also, C++ is very fast and efficient, so it is very useful for intensive and high precision applications such as mix integer programming and other types.
The reason given by Najvot is certainly one, but there are more: 1. You might have a bigger programming project, of which solving an optimization problem is only a small subtask. That big project is written in language XYZ, thus you want to create your model in language XYZ ( and process the results in that language afterwards). 2. You might want to make use of advanced modelling capabilities. E.g., next to your base model you want to add information (like additional constraints) to the model depending on intermediate steps of the optimization process (like individual LP relaxation solutions). Some modelling languages, like Mosel, support use of so-called callbacks, others, like Zimpl, don't. At the latest when you require functionality that is limited to a specific solver, a generic modeling language will probably not support this and you need to use that solver's API.
When you talk about one of the biggest tech companies releasing a free tool for a specific task most probably is a state of the art algorithm or the second best. A preexistent tool or library won't compete with google.
Thanks for the brief overview! As you mentioned that there are some high-level APIs to quickly model scheduling or VRP-like problems, I was wondering if these API's can be used to model a problem that contains both routing and scheduling. An example of such can be home service assignment problem where we need to optimize the total cost to hire servers that serve multiple customers and also schedule their routes such that the customers are served in their preferred time duration. Your thoughts on this would be highly appreciated!
I have to admit Im not a programmer so Im quite laic but I was always wondering why I need these programming tools if I can write my problem by an algebraic modeling language for example AMPL and just "give" my model to a solver. What are the advantages of formulate my problem in C++ or any other programming language?
Because most programming languages are open source and can do what commercial packages can do. Also, C++ is very fast and efficient, so it is very useful for intensive and high precision applications such as mix integer programming and other types.
The reason given by Najvot is certainly one, but there are more:
1. You might have a bigger programming project, of which solving an optimization problem is only a small subtask. That big project is written in language XYZ, thus you want to create your model in language XYZ ( and process the results in that language afterwards).
2. You might want to make use of advanced modelling capabilities. E.g., next to your base model you want to add information (like additional constraints) to the model depending on intermediate steps of the optimization process (like individual LP relaxation solutions). Some modelling languages, like Mosel, support use of so-called callbacks, others, like Zimpl, don't. At the latest when you require functionality that is limited to a specific solver, a generic modeling language will probably not support this and you need to use that solver's API.
When you talk about one of the biggest tech companies releasing a free tool for a specific task most probably is a state of the art algorithm or the second best. A preexistent tool or library won't compete with google.