Learning Forum

[Leonid Pascar]

Hi,

 

As far as I understand from the API when it comes to the parametrized inverse-design the available hyperpartmers relate ot the stopping considitions or the how to define the objective function. Is it true to say that there is no access to parmeters that affect the optimization process such as the update rate?

Thanks in advance,

Leonid

[Taylor Robertson]

Hello Leonid,

I think there is a misunderstanding about what exactly lumopt does. Lumopt is an implementation of the adjoint method, from applied mathemetics, to calculate the gradient of higher dimensional metric spaces - Getting Started with lumopt - Python API. With the gradient you can use Scipy - optimize as a wrapper to get optimal results. There are some parameters to this, but not what I think you are refering to. Gradient descent is an important part of machine learning, but not the complete story. You could concieve of using scikit learn instead of scipy for some additional capabilities, and that would be where you would set the hyperparaemeters.

Best,