Neural network approximation of an inverse functional

The cascade correlation algorithm is used to generate neural networks by learning the inverse of a functional that represents resistivity information of geologic structures. Based on synthetic data several experiments are made to generate and test the neural networks. The generated networks can generalize even when more complex patterns than the used during training are applied. The networks can be used as an internal module in a more general inversion program, or their outputs can be applied to an optimization program if desired. The size of the networks is strongly dependent of the hidden units´ activation function