<div dir="ltr"><div><div><div><div><div><div><div>Dear Ambros, Benjamin, Serrano and Stefan,<br>I address you this question because at <a href="http://scip.zib.de">http://scip.zib.de</a> you are "marked" as developers related to "nonlinear problems".<br><br></div>We have a problem with polynomial constraints of the form (x, y, z - are variables)<br></div>(1) x*y*z*z = 1;<br><br></div>The question is:<br></div>are there reasons (performance or accuracy) to reduce degree of multi-linear term at the cost of additional (auxiliary) variables in nonlinear non-convex continuous problems solved by SCIP ?<br><br></div>E.g. above constraint may be replaced by two constraints of less degree with additional variable v:<br></div>(2) x*y = v; v*z*z = 1;<br></div><div><div><div><div><div><div><div><br></div><div>Preliminary test starts of SCIP 4.* and SCIP 3.2.1 shows that for both variants of the same problem (dozens of variables) results are very close. <br></div><div>The case (1) is solved slightly "faster", but sometimes the following warning appears in the output (problem is solved):</div><div>....<br>[nonlinear] <cons_z[1,2]>: ((((<v[1,1]> + ((-1) * <v[2,1]>)) * (<v[1,1]> + ((-1) * <v[2,1]>))) + ... )))) * (<z[1,2]> * <z[1,2]>)) == 1;<br>violation: left hand side is violated by 1.08...e-06<br><div><div>.....</div><div>The case (2) is solved a little bit longer but without any constraints' violations...</div><div><br></div><div>I took a look at the following report (it was not very deep reading though)<br></div><div>Stefan Vigerske, Ambros Gleixner. SCIP: Global Optimization of Mixed-Integer Nonlinear Programs in a Branch-and-Cut Framework (ZIB Report, May 2016)<br><a href="https://opus4.kobv.de/opus4-zib/files/5937/ZR-16-24.pdf">https://opus4.kobv.de/opus4-zib/files/5937/ZR-16-24.pdf</a></div><div>but couldn't find some explicit recommendations on the subject.</div><div><br></div><div>Are there any unambiguous rules relating to above alternative (1)vs(2)? <br></div><div><br clear="all"></div><div>Regards, <br><div class="gmail_signature"><div dir="ltr"><div>Vladimir V. Voloshinov,</div><div>---------------<br></div><div>Center of Distributed Computing, <a href="http://distcomp.ru" target="_blank">http://distcomp.ru</a>, <br></div><div>Institute for Information Transmission Problems RAS, <a href="http://www.iitp.ru" target="_blank">http://www.iitp.ru</a></div></div></div>
</div></div></div></div></div></div></div></div></div></div>