Topics related to Lumerical and more

Problems with LumOpt- FunctionDefinedPolygon and ParameterizedGeometry

    • Leonid Pascar


      I try using to optimize an array of etched slots and I followed the example of inverse design of grating coupler. I defined the params using FunctionDefinedPolygon or ParameterizedGeometry. In both cases, the inital shape is correct, but the optimization end before the first iteration saying that projg=0. Which is incorrect as the FOM is far from the target. I understand that gradients are not really calculated. Could you please advise about why it happens?


      I pasted below the output of the optimization





      At iterate    0    f= -2.40344D-01    |proj g|=  0.00000D+00

                 * * *

      Tit   = total number of iterations
      Tnf   = total number of function evaluations
      Tnint = total number of segments explored during Cauchy searches
      Skip  = number of BFGS updates skipped
      Nact  = number of active bounds at final generalized Cauchy point
      Projg = norm of the final projected gradient
      F     = final function value

                 * * *

         N    Tit     Tnf  Tnint  Skip  Nact     Projg        F
         17      0      1      0     0     0   0.000D+00  -2.403D-01
        F = -0.24034417377896422

    • Guilin Sun
      Ansys Employee

      My colleague is busy in projects. I just want to suggest to check the optimized parameter and see how does it modify the polygon. You may try a sweep and see if the parameter can actually apply to the polygon.

Viewing 1 reply thread
  • You must be logged in to reply to this topic.