Artificial Neural Networks

We explore the possibilities of using supervised and unsupervised neural networks for the retrieval of microphysical particle properties. First tests with different designs of supervised artificial neural networks (ANN) have just recently been finished. These tests target the question if ANN can provide us with a-priori information that can be used as additional input in our established inversion methods. With regard to unsupervised networks or self-organized mapping (SOM) we investigate structures in large optical data sets which are delivered by remote sensing data.