search

artofscience / SAOR

Sequential Approximate Optimization Repository
GNU General Public License v3.0
5 stars 1 forks source link

Svanberg1987 Two-Bar-Truss without move limits #110

Open artofscience opened 2 years ago

artofscience commented 2 years ago

The test example for the 2bartruss gives exact (up to some digits) the results from the paper. However, upon removal of the move limit strategy, there is deviation. Where is this coming from?

Some things to check:

  1. Move limit from intervening variables (clipping function)
  2. Move limit from MMA (clipping function)
  3. Asymptote update rules
dirkmunro89 commented 2 years ago

I cant reproduce the results for 'without move limits' either (the rest I can reproduce). I tried a number of things, trying to guess / reverse engineer what exactly he did there; have not figured it out yet.