Learning Forum

[achen87]

I try to use LS OPT to search for material parameters but it only runs 3 iterations(the first 2 have 10 points but the last iteration only has 1) the result does not converge to the ground true. Does anyone have any suggestion? thank you!

[igandiko]

Please check if the computed and test histories for the first couple of iterations make sense by plotting them together. If the optimization ran only 1 design at iteration 3, it means it either convereged in two itearitons or there is an inherent issue in the setup. Maybe the computed and test histories are in different units? 

[achen87]

Thank you for your reply.

Here is the first, second and thrid iteration:

 

[igandiko]

How is the error being calculated? Are you using MSE response as the objective? How does the optimization history plot look for the objective and design variables? Perhaps optimum value hit the variable bounds?