<html>
  <head>
    <meta content="text/html; charset=ISO-8859-1"
      http-equiv="Content-Type">
  </head>
  <body bgcolor="#FFFFFF" text="#000000">
    I've seen this as well on all kinds of problems.&nbsp; Our problems have
    a lot of equality constraints, I found by accident that setting
    "bound_relax_factor" to 0.0 to help with equalities also greatly
    reduced changes from run to run.&nbsp; I never chased up why.<br>
    <br>
    Damien<br>
    <br>
    <div class="moz-cite-prefix">On 2014-11-13 4:45 PM, Tony Kelman
      wrote:<br>
    </div>
    <blockquote cite="mid:1603FCA3C83B49A4A96482539F72511D@TKsamsung"
      type="cite">
      <div dir="ltr">
        <div style="FONT-SIZE: 12pt; FONT-FAMILY: 'Calibri'; COLOR:
          #000000">
          <div>The most common case of non-determinism is multithreaded
            BLAS/LAPACK libraries from MKL or OpenBLAS. Those are
            generally not exactly deterministic beyond machine precision
            unless you run them with a single thread. Some of the
            multithreaded sparse linear solvers like MA86 are also
            non-deterministic due to parallelism. If you really want to
            hunt this down, as I said in those linked threads from last
            year, set a high print level and output the logs to
            different files, to compare whether everything is truly
            starting at exactly the same point, and if so, where things
            start differing and by how much. With complicated non-convex
            problems, it&#8217;s entirely possible that initially very minor
            numerical roundoff differences can accumulate and eventually
            lead Ipopt down a different convergence path to a different
            solution.</div>
          <div>&nbsp;</div>
          <div>-Tony</div>
          <div>&nbsp;</div>
          <div style="FONT-SIZE: small; TEXT-DECORATION: none;
            FONT-FAMILY: &quot;Calibri&quot;; FONT-WEIGHT: normal;
            COLOR: #000000; FONT-STYLE: normal; DISPLAY: inline">
            <div style="FONT: 10pt tahoma">
              <div>&nbsp;</div>
              <div style="BACKGROUND: #f5f5f5">
                <div style="font-color: black"><b>From:</b> <a
                    moz-do-not-send="true" title="ipoptuser@gmail.com"
                    href="mailto:ipoptuser@gmail.com">Ipopt User</a> </div>
                <div><b>Sent:</b> Thursday, November 13, 2014 5:35 AM</div>
                <div><b>To:</b> <a moz-do-not-send="true"
                    title="Guy.Charles@redbulltechnology.com"
                    href="mailto:Guy.Charles@redbulltechnology.com">Guy
                    Charles</a> </div>
                <div><b>Cc:</b> <a moz-do-not-send="true"
                    title="ipopt@list.coin-or.org"
                    href="mailto:ipopt@list.coin-or.org">ipopt@list.coin-or.org</a>
                </div>
                <div><b>Subject:</b> Re: [Ipopt] Is ipopt deterministic?</div>
              </div>
            </div>
            <div>&nbsp;</div>
          </div>
          <div style="FONT-SIZE: small; TEXT-DECORATION: none;
            FONT-FAMILY: &quot;Calibri&quot;; FONT-WEIGHT: normal;
            COLOR: #000000; FONT-STYLE: normal; DISPLAY: inline">
            <div dir="ltr">IPOPT itself is fully deterministic, but some
              dependencies may not be deterministic. Please see <a
                moz-do-not-send="true"
                href="http://list.coin-or.org/pipermail/ipopt/2013-April/003337.html">http://list.coin-or.org/pipermail/ipopt/2013-April/003337.html</a>
              and <a moz-do-not-send="true"
href="http://list.coin-or.org/pipermail/ipopt-tickets/2013-March/000635.html">http://list.coin-or.org/pipermail/ipopt-tickets/2013-March/000635.html</a>
              for possible explanations of the behavior you are seeing.
            </div>
            <div class="gmail_extra">
              <div>&nbsp;</div>
              <div class="gmail_quote">On Thu, Nov 13, 2014 at 6:35 AM,
                Guy Charles <span dir="ltr">&lt;<a
                    moz-do-not-send="true"
                    href="mailto:Guy.Charles@redbulltechnology.com"
                    target="_blank">Guy.Charles@redbulltechnology.com</a>&gt;</span>
                wrote:<br>
                <blockquote class="gmail_quote" style="PADDING-LEFT:
                  1ex; MARGIN: 0px 0px 0px 0.8ex; BORDER-LEFT: #ccc 1px
                  solid">
                  <div vlink="#954F72" link="#0563C1" lang="EN-GB">
                    <div>
                      <p class="MsoNormal"><span style="COLOR: #1f497d">Hi,</span></p>
                      <p class="MsoNormal"><span style="COLOR: #1f497d"></span>&nbsp;</p>
                      <p class="MsoNormal"><span style="COLOR: #1f497d">More
                          out of curiosity really:&nbsp; </span></p>
                      <p class="MsoNormal"><span style="COLOR: #1f497d"></span>&nbsp;</p>
                      <p class="MsoNormal"><span style="COLOR: #1f497d">At
                          times I&#8217;m getting some non-deterministic
                          behaviour in my iterations and no of
                          iterations before Ipopt finds an optimum.&nbsp;
                          Despite running the same script twice in a row
                          with identical conditions, I may get different
                          results.&nbsp; I assume that the &#8216;error&#8217; is in my
                          function calls, scripts or layers of code
                          between ipopt and the function calls, but I
                          thought I should double check that Ipopt is
                          completely deterministic.</span></p>
                      <p class="MsoNormal"><span style="COLOR: #1f497d"></span>&nbsp;</p>
                      <p class="MsoNormal"><span style="COLOR: #1f497d">Should
                          ipopt be completely deterministic from each
                          isolated solve call?</span></p>
                      <p class="MsoNormal"><span style="COLOR: #1f497d">Is
                          there any random jitter added in when it
                          encounters an issue?&nbsp; </span></p>
                      <p class="MsoNormal"><span style="COLOR: #1f497d">Would
                          a badly scaled problem cause non-deterministic
                          issues w.r.t numerical precision?&nbsp; </span></p>
                      <p class="MsoNormal"><span style="COLOR: #1f497d">Is
                          it possibly that anything in the linear
                          solvers that could cause this?</span></p>
                      <p class="MsoNormal"><span style="COLOR: #1f497d"></span>&nbsp;</p>
                      <p class="MsoNormal"><span style="COLOR: #1f497d"></span>&nbsp;</p>
                      <p class="MsoNormal"><span style="COLOR: #1f497d">Thanks
                          in advance for any help,</span></p>
                      <p class="MsoNormal"><span style="COLOR: #1f497d"></span>&nbsp;</p>
                      <p class="MsoNormal"><span style="COLOR: #1f497d">Guy</span></p>
                      <p class="MsoNormal"><span style="COLOR: #1f497d"></span>&nbsp;</p>
                      <p class="MsoNormal"><span style="COLOR: #1f497d"></span>&nbsp;</p>
                    </div>
                    <br clear="all">
______________________________________________________________________<br>
                    This email has been scanned by the Symantec Email
                    Security.cloud service.<br>
                    For more information please visit <a
                      moz-do-not-send="true"
                      href="http://www.symanteccloud.com"
                      target="_blank">http://www.symanteccloud.com</a><br>
______________________________________________________________________<br>
                  </div>
                  <br>
                  _______________________________________________<br>
                  Ipopt mailing list<br>
                  <a moz-do-not-send="true"
                    href="mailto:Ipopt@list.coin-or.org">Ipopt@list.coin-or.org</a><br>
                  <a moz-do-not-send="true"
                    href="http://list.coin-or.org/mailman/listinfo/ipopt"
                    target="_blank">http://list.coin-or.org/mailman/listinfo/ipopt</a><br>
                  <br>
                </blockquote>
              </div>
              <div>&nbsp;</div>
            </div>
            <p>
            </p>
            <hr>
            _______________________________________________<br>
            Ipopt mailing list<br>
            <a class="moz-txt-link-abbreviated" href="mailto:Ipopt@list.coin-or.org">Ipopt@list.coin-or.org</a><br>
            <a class="moz-txt-link-freetext" href="http://list.coin-or.org/mailman/listinfo/ipopt">http://list.coin-or.org/mailman/listinfo/ipopt</a><br>
          </div>
        </div>
      </div>
      <br>
      <fieldset class="mimeAttachmentHeader"></fieldset>
      <br>
      <pre wrap="">_______________________________________________
Ipopt mailing list
<a class="moz-txt-link-abbreviated" href="mailto:Ipopt@list.coin-or.org">Ipopt@list.coin-or.org</a>
<a class="moz-txt-link-freetext" href="http://list.coin-or.org/mailman/listinfo/ipopt">http://list.coin-or.org/mailman/listinfo/ipopt</a>
</pre>
    </blockquote>
    <br>
  </body>
</html>