Abstract
We present a discrete and asynchronous implementation of Vidyasagar mean-field neural net for the identification of faulty elements in large antenna arrays, from field or intensity measurements. The underlying factorially complex, non convex minimization problem is introduced, and the limitations of standard approaches are reviewed. The convergency properties of the net are discussed, together with several implementation issues. Extensive numerical experiments show that the proposed algorithm is reliable, robust and fast, and overall superior to standard (conjugate-gradient based) techniques.