DESIGN OF ARTIFICIAL NEURAL NETWORK FOR SOLVING INVERSE PROBLEMS

Abstract

This paper proposes neural network based forward models in iterative inversion algorithms for solving inverse problems .Iterative algorithms are commonly used to solve inverse problem.Typical iterative inversion approaches use a numerical forward model to predict the signal for a given input data. The desired output can then be found by iteratively minimizing energy function. The use of numerical models is computationally expensive, and therefore, alternative forward models need to be explored. This study proposes two different neural network based iterative inverse problem solutions. In addition, specialized neural networks forward models are proposed and used in place of numerical forward models. The first approach uses basis function networks (radial basis function (RBFNN)) to approximate the mapping from the input space to the output space. The back propagation training algorithm are then used to estimate the network parameter. The second approach proposes the use of two networks in a feedback configuration.