Abstract-In this paper, we investigate nonuniformly spaced linear arrays (NUSLA) rigorously. Several important problems in NUSLA design are solved with the combination of the Genetic Algorithm and Conjugate Gradient method (GA-CG). The pattern synthesis for the specified beamwidth and minimum achievable sidelobe level (SLL) are performed and for the first time, the graphs which show the relation between the beamwidth, sidelobe level and number of elements for NUSLA are derived. Also, the NUSLA's pattern for the specified directivity and sidelobe level is synthesized. The graphs showing the behavior of NUSLA relative to the increase of its length for a fixed number of elements are derived. These graphs show the relations between the directivity and sidelobe level of NUSLA with its length. As a practical design, an array of parallel dipoles is designed for specified beamwidth/sidelobe level or specified directivity/sidelobe level. Furthermore, a novel Neural Network based model for the NUSLA is presented for the rapid and accurate computation of Sparameters. The computed S-parameters are used for the computation of coupling among elements. Then the GA-CG method can adjust these values in the synthesis process to achieve desired pattern and bearable coupling among elements.