Attached is a very simple "no frills" matlab SA code for optimizing
functions with real valued inputs that I use in my computational intelligence course for educational purposes, not anywhere near the sophistication of Robert's problem where you had discrete inputs and all sorts of complex constraints. In my experience, if you are working with smooth objective functions involving real valued inputs, adaptive simulated annealing works really well (http://www.ingber.com/#ASA). There is fun demo (though not TSP based) at http://exatech.com/Optimization/Optimization.htm There was also a very readable amateur scientist article in the March 1997 issue of Scientific American on SA. It was on archive.org (http://web.archive.org/web/*/http://www.sciam.com/0397issue/0397amsci.html), but I noticed that they have removed it. ____________________________________________________ [hidden email] http://www.optomatica.com mobile: +20 (10) 191 6650 ----- Original Message ----- From: "Robert Holmes" <[hidden email]> To: "'The Friday Morning Applied Complexity Coffee Group'" <[hidden email]> Sent: Tuesday, October 12, 2004 12:59 AM Subject: RE: [FRIAM] Simulated Annealing >I used SA in a real-life application, optimizing the allocation of staff to > shifts on a power station. An interesting problem, because as well as > having > to deal with the getting the right grades of people on any given shift > (unit > operator, assistant unit operator, maintenance etc.), you had to get the > right mix of skills (electrician, engineer etc.) and make sure that no > temporal constraints were being broken (you can only work the night shift > three nights in a row, everyone needs to have done the same number of > day-shifts/night-shifts etc over a year). Add into this the need to > allocate > overtime fairly and deal with unplanned absences, and things got a little > complex. > > The solution: represent this all as one big-ass objective function and let > a > simple SA program lose on it. Usually got a good enough solution within 5 > minutes or so. I got the SA algorithm from Numerical Recipes. > > - rh > >> -----Original Message----- >> From: Owen Densmore [mailto:[hidden email]] >> Sent: Monday, October 11, 2004 4:25 PM >> To: The Friday Morning Applied Complexity Friam >> Subject: [FRIAM] Simulated Annealing >> >> Does anyone have a pointer to using simulated annealing to >> solve the traveling salesman problem? .. or a good tutorial on SA? >> >> -- Owen >> >> Owen Densmore - http://backspaces.net - [hidden email] >> >> >> ============================================================ >> FRIAM Applied Complexity Group listserv >> Meets Fridays 9AM @ Jane's Cafe >> Lecture schedule, archives, unsubscribe, etc.: >> http://www.friam.org >> >> >> > > > > ============================================================ > FRIAM Applied Complexity Group listserv > Meets Fridays 9AM @ Jane's Cafe > Lecture schedule, archives, unsubscribe, etc.: > http://www.friam.org > > A non-text attachment was scrubbed... Name: SAmin.m Type: application/octet-stream Size: 1490 bytes Desc: not available Url : /pipermail/friam_redfish.com/attachments/20041012/629ff797/SAmin.obj |
Free forum by Nabble | Edit this page |