Előadás címe: Localization of scattering object using neural networks
Helyszín: Online
Időpont: 2020.06.30., 10:15 – 10:35
Kivonat: The inverse scattering problem is investigated in two space dimensions. A real life-situation is chosen, where the scattered wave is known only on a segment instead of having a far field pattern. Also, we have two scattering bodies, which have to be deteced. For this, a neural network is applied, where the training data is generated using a MATLAB simulation package. The structure of the neural network is explained and displayed in details, regarding the number of layers, activation functions, learning rates and dropouts. The construction is implemented in Python. The simulation results are accurate: using the validation data, we could estimate the location of the scattering objects with a good precision.