The question about get_max_pairing() about auctionassignment

[Azmisov]

In the example in README, there are two user defined functions I didn't show:

1. get_pairing_cost: The arc benefit for matching a person and object
2. get_max_pairing: This is the maximum value out of all of the arc benefits you specify in #1; e.g. out of all the possible person-object pairs, which has the max cost? This is used internally as an upper bound when doing the iterative solve. I could have had the solver look through the arc matrix itself and find the max value, but for my use case, I knew what the max would be analytically; so the solver has you calculate and pass it in yourself.

Does that clarify it?

[wangshui123]

Thank you very much. I understand what you mean. Can you leave me an email address so that I can ask you further questions. Thank you again! !

…

------------------&nbsp;原始邮件&nbsp;------------------ 发件人: "Azmisov/AuctionAssignment" <[email protected]&gt;; 发送时间:&nbsp;2020年12月26日(星期六) 上午10:12 收件人:&nbsp;"Azmisov/AuctionAssignment"<[email protected]&gt;; 抄送:&nbsp;"小冢虎王水"<[email protected]&gt;;"Author"<[email protected]&gt;; 主题:&nbsp;Re: [Azmisov/AuctionAssignment] The question about get_max_pairing() (#1) In the example in README, there are two user defined functions I didn't show: get_pairing_cost: The arc benefit for matching a person and object get_max_pairing: This is the maximum value out of all of the arc benefits you specify in #1; e.g. out of all the possible person-object pairs, which has the max cost? This is used internally as an upper bound when doing the iterative solve. I could have had the solver look through the arc matrix itself and find the max value, but for my use case, I knew what the max would be analytically; so the solver has you calculate and pass it in yourself. Does that clarify it? — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.

[wangshui123]

hello， I have another important question，thank you.
When I want to match m persons with n objects (m<n), can the algorithm ensure that all n objects can be assigned to different persons ?
Or is the algorithm only capable of matching m objects among n objects to m persons to maximize benefits?

[Azmisov]

This is for the "assignment problem", which will only assign one person to one object and vice versa. So yes, they will all be different.

The case where an object can be assigned to multiple persons is a different problem and needs a different algorithm. You can look into "weapon target assignment problem" if you're interested. There's also the generalized assignment problem too. The less constrained the problem gets, in general the algorithms to solve it will be slower too; so just be aware of that.

Also sorry, I'd prefer not to share my email here if that's okay... Better not to have it out in the open to reduce spam and other stuff :)

[wangshui123]

Thank you very much, I understand what you mean. If there are some issues that need to be discussed in the future, I hope to continue to ask you on github. Thank you again! ! Thank you