Tolerance and constraint parameters

Answered

Comments

3 comments

  • Yuriy Zinchenko

    Hello Soufyan:

    the names "interior-point" and "barrier" are often synonymous when referring to optkmizaiotn algorithms, although I cannot guarantee that this is exactly what Matlab's solver is using.

    If you are solving QP with Gurobi's barrier, you may want to look into

    https://www.gurobi.com/documentation/9.0/refman/barqcpconvtol.html

    as this parameter is used to determine the convergence criterion for our barrier.

    For models where the data structures are sparse, it is often (very) advantageous to specify those as such, without the artificial step of making a full matrix filled with many 0's; indeed, linear algebra --a driving engine, so to speak-- behind the barrier is often more efficient with sparse matrices.

     

    Hope this helps,

     

    0
    Comment actions Permalink
  • Soufyan Zayou

    Hello Yuriy,

    Thanks for your reply. 

    I don't really understand the last sentence of your answer: ''indeed, linear algebra --a driving engine, so to speak-- behind the barrier is often more efficient with sparse matrices.''

    Could you elaborate more on this?

    Furthermore, if I understand your answer correctly the data structures that Gurobi uses, are sparse. And due to that, it can handle sparse matrices more easy.

    Regards,

    0
    Comment actions Permalink
  • Yuriy Zinchenko

    At every iteration of the barrier one solves a system of linear equations, and this is often the most expensive part.

     

    >> the data structures that Gurobi uses, are sparse

    yes, you can think of it this way.

     

    Hope this helps.

     

     

    0
    Comment actions Permalink

Please sign in to leave a comment.

Powered by Zendesk