[SCIP] Time since last improvement

[Mexi, Gioni]

Hi Eliel,


I am not aware of such a parameter in SCIP. The closest one is the parameter:


limits/stallnodes          solving stops, if the given number of nodes was processed since the last improvement of the primal solution value (-1: no limit) [-1]

Here are some other SCIP parameters that may enable a fair comparison of the two solvers.

  limits/absgap                solving stops, if the absolute gap = |primalbound - dualbound| is below the given value [0.0]
  limits/bestsol               solving stops, if the given number of solution improvements were found (-1: no limit) [-1]
  limits/gap                      solving stops, if the relative gap = |primal - dual|/MIN(|dual|,|primal|) is below the given value, the gap is 'Infinity', if primal and dual bound have opposite signs [0.0]
  limits/nodes                 maximal number of nodes to process (-1: no limit) [-1]
  limits/objective            set limit on objective function, such that only solutions better than this limit are accepted
  limits/softtime             soft time limit which should be applied after first solution was found (-1.0: disabled) [-1.0]
  limits/solutions            solving stops, if the given number of solutions were found (-1: no limit) [-1]
  limits/time                   maximal time in seconds to run [1e+20]
  limits/totalnodes        maximal number of total nodes (incl. restarts) to process (-1: no limit) [-1]

I hope this helps.

Best regards,

Gioni
________________________________
Von: Scip <scip-bounces at zib.de> im Auftrag von Eliel Dzik <edzik at wiremind.io>
Gesendet: Mittwoch, 9. August 2023 09:14:07
An: scip at zib.de
Betreff: [SCIP] Time since last improvement


Hello,

I am a SCIP user and I'm wondering if there is a parameter or a way to set a stopping condition for when x seconds have passed since the last improvement of the solution, even if the current solution is not optimal. I know that other solvers have implemented this feature, and I have searched forums for information on how to do this in SCIP but have not found any.

My task is to compare two solvers, and SCIP is one of them. Although it is often faster than its competitor, for most complex cases it is worse because it waits until reaching the time limit, whereas its competitor can stop earlier.

Thanks for your helpful documentation, I hope you will be able to answer my questions.

Best regards,

Eliel DZIK





[https://lh5.googleusercontent.com/CKjVeXH2mtXHU4qqQrytaKvrRlqAmhauluIbkNNqjyvfW477QCEDYK8Q1BEmuYoATwEFimu5gK4dZ8csz50cZ_2NXh_7ChbhCQ_yWz7Z4eAW8ozT9sMruCMl7OcQMImUffWwmEjC4KDIMBTK7WHJxg]

  [https://lh5.googleusercontent.com/GBNehRKiCKG47Nb2ib0fG1tHPQcKfi2aaxCVvTESwU6n6I9ZpeCGvfKdD7lESWTGa-Ehool5CxDPByC8L7-PmLyvUjriYWXlGivVswW0Ejem-QlEscYcJwtPJbmHaSWX1z9aoZUhueG-Jq7GGu8GfQ] <https://www.wiremind.io/>  [https://lh3.googleusercontent.com/3y8QkWTC2yHwD3mkmxPJYRkEG4bjqtCSRqC9fKr9ZNVfx8MtAbg2pXf7AvKsrqKI-oDZggBXAyuxKoM39hMOH_KvhGkPMi6mO9h6FF8zm_oh2ug2tyGej_pGoXSTa1_gAl_DBaczNmPcdF5GYE7H0w] <https://www.linkedin.com/company/wiremind>





Eliel DZIK

Optimization Engineer

edzik at wiremind.io<mailto:edzik at wiremind.io>
<mailto:cgiraultmatz at wiremind.io>


16, boulevard Poissonnière - 75009 Paris, France




The content of this email is confidential and intended for the recipient specified in message only.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.zib.de/pipermail/scip/attachments/20230809/65be495f/attachment.html>