In real world engineering applications the uncertainties of the structural parameters are inherent and the scatter from their nominal ideal values is in most cases unavoidable. These uncertainties play a dominant role in structural performance and the only way to assess this influence is to perform Reliability-Based Design Optimization (RBDO) and Robust Design Optimization (RDO). Compared to the basic deterministic-based optimization problem, a RBDO problem considers additional non-deterministic constraint functions, while the RDO yields a design with a state of robustness, so that its performance is the least sensitive to the variability of the uncertain parameters. The first part of this study examines the application of Neural Networks (NN) to the RBDO of large-scale structural systems, while the second part investigates the structural RDO problem. The use of NN in the framework of the RBDO problem is motivated by the approximate concepts inherent in reliability analysis and the time-consuming repeated analyses required by Monte Carlo Simulation. On the other hand the RDO is a multi-criteria optimization problem where the aim is to minimize both the weight of the structure and the variance of the structural response.