Using deep-learning methods to invers design metamaterials structures 

To generate the data i consider 6 parameters, thickness of 3 bars of MoS (L1, L2, L3) (molybdenum de sulfat) (see: en.wikipedia.org/wiki/Molybdenum_disulfide for more information). and the carrier densities of these three bars (N1, N2, N3) (see: ieeexplore.ieee.org/document/7762817 for more information

Precisely trying to understand the data collection, data processing parts.

I have seen your paper also.

@@shahnawaz27790 CST simulations he said

to generate your dataset for training your Network you can use any simulator, CST, COMSOL, HFSS, etc. and use parameter sweep. However, I suggest you choose one with good accuracy and low simulation time. since generating data take lots of time